No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Optimization of ATM cash replenishment with group-demand forecast
Abstract
In ATM cash replenishment banks want to use less resources (e.g., cash kept in ATMs, trucks for loading cash) for meeting fluctuated customer demands. Traditionally, forecasting procedures such as exponentially weighted moving average are applied to daily cash withdraws for individual ATMs. Then, the forecasted results are provided to optimization models for deciding the amount of cash and the trucking logistics schedules for replenishing cash to all ATMs. For some situations where individual ATM withdraws have so much variations (e.g., data collected from Istanbul ATMs) the traditional approaches do not work well. This article proposes grouping ATMs into nearby-location clusters and also optimizing the aggregates of daily cash withdraws (e.g., replenish every week instead of every day) in the forecasting process. Example studies show that this integrated forecasting and optimization procedure performs better for an objective in minimizing costs of replenishing cash, cash-interest charge and potential customer dissatisfaction.
... In [7] proposes the optimization of service costs using a predictive cash replenishment strategy for an ATM network in a Russian bank, the strategy consists of three stages: 1) Data processing and extraction of characteristics: ATM identifier, daily cash flow, daily ATM replenishment, date, location, fixed costs; 2) Evaluation of the predictive model, here linear models and non-linear automatic learning models are evaluated; and 3) Money replenishment plan, in which decisions are made based on the data obtained in the previous stage. In [8] proposes the optimization of the replenishment of money in ATMs with forecasting of the demand by ATM groups, this through a methodology that includes grouping ATMs by their proximity and then forecasting demand by groups, thereby optimizing costs in relation to the models that forecast withdrawals at ATMs individually posed in [9]. These works were carried out on information provided by the "Forecast Contest for Artificial Neural Networks and Computational Intelligence" (NN5 Competition), which offered data on cash withdrawals from an ATM network in the United Kingdom, and consists of forecasting a set of 11 or 111 daily time series of cash withdrawals at ATMs with the highest possible precision, using computational intelligence methods and effective methodology [10]. ...
... In [14] proposes a NN as an alternative for monitoring the operation of the conditions of a hydro generator and these results are used to carry out a predictive maintenance plan; which is precisely what it is want to achieve in this study. In this paper, a methodology similar to that proposed in [7], but adding the grouping of ATMs by common characteristics and proximity, how a valid alternative to improve the results of the forecasts [8]. With these results, define a maintenance plan and optimize the routes that technicians must travel, making an analogy to the Traveling Salesman Problem (TSP). ...
... The objective function is the function that tells how good a path is, in this case it is the sum of the distances of a path. The objective function for the TSP defined in [19] is shown in (5), under the constrains (6),(7), (8) and (9). ...
... Aiming to maximize income from transactions and satisfy customer demand for cash, some banks might store as much as 40% more banknotes in ATMs than they actually need [38]. However, loading excess cash in ATMs, rather than only loading in what the demand roughly is, will increase operational and opportunity costs [9,20,21]. Conversely, if there is not enough cash loaded into ATMs, there will be "out of cash" transactions, resulting in the bank's reputation being damaged, as well as lowered income and customer satisfaction. Thus, a more accurate prediction of ATM currency demand can help financial institutions avoid being tempted to fill ATMs with too many notes and earn more profit by mobilizing idle cash and generating additional revenue through investments-specifically in countries with high-interest rates and overnight interest rates. ...
... However, some significant features, such as the number of consecutive holidays ahead, have not been included in the previous studies. Besides, only a few papers considered both time-and location-related variables (e.g., [20,21]), though the location of ATMs can meaningfully affect the amount of daily cash withdrawn from these ATMs. For instance, Ekinci et al. [20] included a location feature as an independent variable of the model and proposed grouping ATMs into nearby-location clusters. ...
... Besides, only a few papers considered both time-and location-related variables (e.g., [20,21]), though the location of ATMs can meaningfully affect the amount of daily cash withdrawn from these ATMs. For instance, Ekinci et al. [20] included a location feature as an independent variable of the model and proposed grouping ATMs into nearby-location clusters. However, using a more meticulous outlook, the nearby-location and/or the same geographical location of ATMs do not necessarily indicate a similar withdrawal pattern, since the points of interest in the ATM's vicinity might be different. ...
The overarching goal of this paper is to accurately forecast ATM cash demand for periods both before and during the COVID-19 pandemic. To achieve this, first, ATMs are categorized based on accessibility and surrounding environmental factors that significantly affect the cash withdrawal pattern. Then, several statistical and machine learning models under different algorithms and strategies are employed. In aiming to provide the feature matrix for machine learning models, some new influential variables are added to the literature. Finally, a modified fitness measure is proposed for the first time to correctly choose the most promising model by considering both the prediction errors and accuracy of direction's change simultaneously. The results obtained by a comprehensive analysis-a statistical analysis together with grid search and k-fold cross-validation techniques-reveal that (i) category-wise prediction enhances forecasting quality; (ii) before COVID-19 and in times when there are only minor disturbances in withdrawal patterns, forecasting quality is higher, and in general, the machine learning models can more appropriately forecast ATM's cash demand; (iii) despite studies in the literature, sophisticated models will not always outperform simpler models. It is found that during COVID-19 and in times when there is a sudden shock in demand and massive volatility in withdrawal patterns, the statistical models of the autoregressive integrated moving average (ARIMA) and seasonal ARIMA (SARIMA) can mainly provide better forecasting likely due to high performance of such models for short-term prediction, while minimizing overfitting.
Supplementary information:
The online version contains supplementary material available at 10.1007/s42979-021-01000-0.
... Although both forecasting techniques and cash replenishment policy optimization models are required to optimize the physical cash in a bank's network, there are a limited number of papers that integrate these two types of models. To the best of our knowledge, there are only five studies that integrate cash demand forecasts into the replenishment policy optimization problem: (Simutis, Dilijonas, Bastina, Friman, & Drobinov, 2007), (Baker, Jayaraman, & Ashley, 2013), (Osorio & Toro, 2012), (Ekinci, Lu, & Duman, 2015) and (Lazaro, Jimenez, & Takeda, 2018). Regarding real-world scenarios, the literature is notably limited. ...
... Regarding real-world scenarios, the literature is notably limited. Although there are certain models, such as (Ekinci, Lu, & Duman, 2015), (Agoston, Benedek, & Gilanyi, 2016), (Lazaro, Jimenez, & Takeda, 2018) and (Osorio & Toro, 2012), The model was later applied to optimize the cash network of a Colombian financial institution that had 85 offices, including branches and several ATMs. ...
... is generally used as the primary performance measure, ranging between 20 % and 45 % (Ekinci, Lu, & Duman, 2015). After the model was implemented in the network, the six months' forecast and the actual cash requirements data for the ATMs in the network (330 ATMs at the time) ...
In this study, an integrated cash requirement forecasting and cash inventory optimization model is implemented in both the branch and automated teller machine (ATM) networks of a mid-sized bank in Turkey to optimize the bank’s cash supply chain. The implemented model’s objective is to minimize the idle cash levels at both branches and ATMs without decreasing the customer service level (CSL) by providing the correct amount of cash at the correct location and time. To the best of our knowledge, the model is the first integrated model in the literature to be applied to both ATMs and branches simultaneously. The results demonstrated that the integrated model dramatically decreased the idle cash levels at both branches and ATMs without degrading the availability of cash and hence customer satisfaction. An in-depth analysis of the results also indicated that the results were more remarkable for branches. The results also demonstrated that the utilization of various seasonal indices plays a very critical role in the forecasting of cash requirements for a bank. Another unique feature of the study is that the model is the first to include the recycling feature of ATMs. The results demonstrated that as a result of the inclusion of the deliberate seasonal indices in the forecasting model, the integrated cash optimization models can be used to estimate the cash requirements of recycling ATMs.
... For the first step, we assume that a reliable forecast for the amount of cash to be withdrawn each day for a period of time (typically for a week) is available. There are several works focused on this first phase of the problem [1], [9], [6], [10], [7]. ...
... While finding the minimum cost, there are two options to be compared: loading cash day by day or at once. To calculate day by day replenishment cost, sum of c [1,1] and c [2,2] can be used as shown in Equation 6. ...
... This work can be further extended by including daily withdrawal predictions. Another line of research direction can be considering ATM groups as in [6]. ...
Automated Telling Machine (ATM) replenishment is a well-known problem in banking industry. Banks aim to improve customer satisfaction by reducing the number of out-of-cash ATMs and duration of out-of-cash status. On the other hand, they want to reduce the cost of cash replenishment, also. The problem conventionally has two components: forecasting ATM cash withdrawals, and then cash replenishment optimization on the basis of the forecast. In this work, for the first component, it is assumed that reliable forecasts are already obtained for the amount of cash needed in ATMs. We focus on the ATM cash replenishment component, and propose a dynamic programming based solution. Experiments conducted on real data reveal that the solutions of the baseline approaches have high cost, and the proposed algorithm can find optimized solutions under the given forecasts.
... In the proposed model, the optimal cash inventory level and the time between orders are obtained via an optimization module, which uses the rolling horizon time series forecasts of cash withdrawals. Ekinci, Lu, & Duman (2015) integrated the results of their forecasts with a cash replenishment optimization model to determine the amount of cash and trucking logistics schedules necessary for replenishing cash in all of the ATMs. Agoston, Benedek, & Gilanyi (2016) defined the cash management problem as a single problem by incorporating the cash optimization problem of the bank into the cash optimization problem of the CIT firms. ...
... The authors advocated that ATMs with similar weekday cash demand patterns generated better results than individual ATM forecasts. Ekinci et al. (2015) grouped ATMs in nearby locations into clusters and generated forecasts for the clusters. The researchers also studied location variables to improve the forecasting model quality. ...
... Although both forecasting techniques and cash replenishment policy optimization models are required to optimize the physical cash in a bank's network, there are a limited number of papers that integrate these two types of models. To the best of our knowledge, there are only five studies that integrate cash demand forecasts into the replenishment policy optimization problem: (Simutis, Dilijonas, Bastina, Friman, & Drobinov, 2007), (Baker, Jayaraman, & Ashley, 2013), (Osorio & Toro, 2012), (Ekinci, Lu, & Duman, 2015) and (Lazaro, Jimenez, & Takeda, 2018). Regarding real-world scenarios, the literature is notably limited. ...
In this study, an integrated cash requirement forecasting and cash inventory optimization model is implemented in both the branch and automated teller machine (ATM) networks of a mid-sized bank in Turkey to optimize the bank’s cash supply chain. The implemented model’s objective is to minimize the idle cash levels at both branches and ATMs without decreasing the customer service level (CSL) by providing the correct amount of cash at the correct location and time. To the best of our knowledge, the model is the first integrated model in the literature to be applied to both ATMs and branches simultaneously. The results demonstrated that the integrated model dramatically decreased the idle cash levels at both branches and ATMs without degrading the availability of cash and hence customer satisfaction. An in-depth analysis of the results also indicated that the results were more remarkable for branches. The results also demonstrated that the utilization of various seasonal indices plays a very critical role in the forecasting of cash requirements for a bank. Another unique feature of the study is that the model is the first to include the recycling feature of ATMs. The results demonstrated that as a result of the inclusion of the deliberate seasonal indices in the forecasting model, the integrated cash optimization models can be used to estimate the cash requirements of recycling ATMs.
... Finally, we consider stock-out as a cost associated with the fail to satisfy the requested cash amount. In the literature, there are several studies considered the problem of joining inventory management and routing problems (Ekinci et al., 2015). ...
... In this study, we consider the forecasted demand for the requested cash at ATM as the number of bills of a particular class being withdrawn form an ATM at a certain period (Ekinci et al., 2015;Venkatesh et al., 2014). More elaboration of the lost-demand will be discussed in following sections. ...
Automated Teller Machines (ATMs) are one of the most important cash distribution channels for the banks. Banks are expanding their ATM network to satisfy customer demand and are facing operating costs. This work focuses on the inventory management for optimizing the replenishment policy of the Automated Teller Machine (ATM) network. We introduce a multi-period strategy that helps to minimize the total cost of ATMs replenishment. In addition, the paper discusses how to obtain the shortest tour to replenish ATMs. We formulate the problem as a Mixed Integer Nonlinear Programming (MINLP) model and solve it using GAMS solver BARON, to minimize the total cost associated with replenishing ATMs. Then, we linearize the MINLP and compare the results of both models.
... Considering the Istanbul ATMs, it has been identified grouping nearby ATM into nearby location clusters and forecasting cash amounts is more practical than forecasting for each ATM separately [4]. ...
... The stationary behavior of the 4 th series is confirmed through "augmented dickey fuller unit root test". By analyzing several time-series ARMA models, ARIMA (1,4,3) is selected as the best-fitted model to forecast ATM cash withdrawal fluctuations for the selected ATM based on statistical measures. R-squared value for the selected model is 97.46% and adjusted R-squared value is 97.44%. ...
... For ATM cash replenishment banks want to use fewer resources and also meet fluctuating customer demands [21,22]. Till now exponential weighted moving average procedure was used to model daily cash replenishment for individual ATMs. ...
... The ATM cash replenishment problem addressed in [21,22] but not in ESObased approach. These research articles provide a basis for the objective function formulation and the application of ESO techniques. ...
Banking industry is the backbone of the economy of any country, and it does have many operational issues as well as other financial issues. As regards to solving operational issues such as Portfolio optimization, Bankruptcy prediction, FOREX rate prediction, ATM cash replenishment, ATM/Branch location prediction, Interbank payments, liquidity prediction, etc., banking industries are moving away from conventional ways toward more automated and more robust methods. Evolutionary and Swarm Optimization (ESO) based techniques play a vital role in solving the above-mentioned operational issues because they yield global or near-global optimal results. We survey most of the works reported in this space starting from 1998 to 2016. While the application of ESO techniques to solve the business issues is well-documented, the same on the operational issues is very relevant.
... Previous research on grouping ATMs into nearby location clusters and optimizing the aggregates of daily cash withdrawals in the forecasting process [5] was proposed. Example studies showed that this procedure was superior as a minimization strategy for costs and replenishments to ensure customer satisfaction. ...
A set of ATMs in commercial areas were considered for business transactions. Customer demand for money requires prompt cash supply; to harmonize optimality options for ATM cash loading to sustain customers. We formulated this tis decision as a multi-stage stochastic process with Markovian demand of ATM cash among customers. The loading and operations of the ATMs were governed by cash demand and supply; with unit (loading, operational, and shortage) costs. The study considered the ATM cash demand of ten ATMs in regions within commercial centres of Kampala city. The relevant data was collected every hour throughout four months. The task focused on optimizing cash-loading decisions that minimize the operational costs of managing ATMs. The optimality of cash loading decisions were determined using the Markov decision processes. The tested model revealed optimal state-dependent options and operational costs of managing ATMs cost-effectively. As a cost optimization strategy for managers in the banking industry, the productivity of ATMs towards improved efficiency and utilization of resources for cash provision to customers emerged; supporting a larger customer base for business transactions.
... To prevent the ATMs from running out of cash, managers often decide to deliver and store significantly more cash in a device than usually needed (Simutis et al., 2007a). However, such a strategy simply leads to higher operating costs that result from the need to prepare the cash, "freeze" it, and transport it (Batı and Gözüpek, 2017;Ekinci et al., 2015Ekinci et al., , 2021. Those activities that generate costs that are related to cash management (which mainly consist of the preparation and transport of cash, "freezing" the cash, and returning any undisbursed cash) may account to up to 50 % of the total costs of operating an ATM network (Simutis et al., 2007a;Suder, 2015;Arnfield, 2017). ...
This study aims to test the forecasting accuracy of recently implemented econometric tools as compared to the forecasting accuracy of widely used traditional models when predicting cash demand at ATMs. It also aims to verify whether the pandemic-driven change in market conditions impacted the predictive power of the tested models. Our conclusions were derived based on a data set that consisted of daily withdrawals from 61 ATMs ofone of the largest European ATM networks operating in Krakow, Poland, and covered the period between January 2017 and April 2021.
The results proved that the recently implemented methods of forecasting ATM withdrawals were more ac-curate as compared to the traditional ones, with XGBoost providing the best forecasts in the majority of the tested cases. Moreover, it was found that the pandemic-driven change in market conditions affected the predictive power of the models. Both of these results seem particularly useful for improving the efficiency of ATM networks.
50 days' free access to article: https://authors.elsevier.com/a/1iLAD98SGwii2
... More recent articles on the use of neural networks in forecasting withdrawals from ATMs are by Serengil and Ozpinar (2019), Ekinci et al. (2015), Zandevakili and Javanmard (2014), Bhandari and Gill (2016). The promising computer-supported methods with respect to ATM applications may be based on evolutionary algorithms which are thoroughly presented in Sieja and Wach (2019). ...
Abstract
Objective: The study focused on verifying the impact of the calendar and seasonal effects on the accuracy of forecasts of cash withdrawals from automated teller machines (ATMs). In this article, we investigated a possible use of the so-called trigonometric seasonality, the Box-Cox transformation, ARMA errors, trend, and seasonal components (TBATS) models to forecast withdrawals from ATMs. In practice, the SARIMA model is widely used as a forecasting tool. However, the major limitation of SARIMA models is that it allows just one single seasonality pattern to be taken into account, e.g., weekly seasonality. At the same time, cash withdrawals from ATMs display overlapping multi-seasonality. Therefore, the goal of this article is to compare the SARIMA model with the TBATS model, both in basic forms and forms extended with event-specific dummies.
Research Design & Methods: Empirical research was conducted by means of fitting SARIMA and TBATS models to daily time series of withdrawals from 74 ATMs managed by one of the largest ATM operators in Poland. The dataset covered the period of 2017–2019.
Findings: Forecasting levels of cash withdrawals plays a crucial role in the management of ATM networks, both in the case of a single ATM as well as the whole network. Prediction accuracy has a direct impact on the operational costs of the network. These costs result from activities such as freezing cash in an ATM, preparing it, and transporting it to an ATM. Therefore, the choice of a proper forecast model is of special importance. According to statistical evidence in our study, the basic TBATS model gives more accurate forecasts than the basic SARIMA model widely used in practice.
Implications & Recommendations: The multi-seasonality of ATM withdrawals means that it is necessary to use techniques that take such phenomena into account in a single joint model. Multi-seasonality can be modelled using TBATS models. The study confirmed that TBATS models can be considered useful alternatives in planning cash replenishments in ATM networks.
Contribution & Value Added: This article is an extensive empirical study on the selection of proper methods and forecasting models necessary to predict withdrawals from ATMs with overlapping multi-seasonalities and calendar effects. We proved that taking seasonal and calendar effects into account when forecasting withdrawals from ATMs significantly reduces forecast errors. Statistically significant improvement in forecast accuracy was observed both for SARIMA and TBATS. After taking calendar effects into account, TBATS forecast errors were slightly smaller than those resulting from corresponding SARIMA models. However, this result is statistically insignificant. The results of this study imply a need for further studies on the applications of TBATS models in forecasting the required cash level in ATMs, which in turn may help improve the efficiency of ATMs network management.
... Information regarding whether the day corresponds to a working day or a weekend is relevant to the problem under study. Ekinci et al. (2015) considered the problem of optimizing ATM cash replenishment in Istanbul using simple linear regression with grouping. They have achieved a MAPE score of 22.69% on average. ...
The problem of forecasting time series is very widely debated. In recent years, machine learning algorithms have been very prolific in this area. This paper describes a systematic approach to building a machine learning predictive model for solving optimization problems in the banking sector. A literature analysis on applying such methods in this particular area is presented. As a direct result of the described research, a universal scenario for forecasting various non-stationary time series in automatic mode was developed. The developed scenario for solving specific banking tasks to improve business efficiency, including optimizing demand for ATMs, forecasting the load on the call center and cash center, is considered. A machine learning methodology in economics that can yield robust and reproducible results and can be reused in solving other similar tasks is described. The methodology described in the article was tested on three cases and showed the ability to generate models that are superior in accuracy to similar predictive models described in the literature by at least three percentage points. This article will be helpful to specialists dealing with the problem of forecasting economic time series and students and researchers due to a large number of links to systematic literature reviews on this topic.
Citation: Gorodetskaya, O.; Gobareva, Y.; Koroteev, M. A Machine Learning Pipeline for Forecasting Time Series in the Banking Sector. Economies 2021, 9, 205. https://doi.org/10.3390/economies9040205
... Whereas these two algorithms are useful a single time series analysis; for multiple time series directly proportional to each other that go hand in hand, it was observed to have used the Vector Auto Regression Moving Average model for analyzing trends and seasonality that effect from one-time series to another. If not making a prediction on a daily basis the data would be less varying and complex models would not be required hence, even Linear Regression would have been a fit for the problem [14]. In some special cases using deep learning techniques for very short time horizon interval (10 minutes, 30 minutes to 1 hour) is consolidated to outperform others, specifically the LSTM neural network. ...
Automatic Teller Machine (ATM) are still largely used to dispense cash to the customers. ATM cash replenishment is a process of refilling ATM machine with a specific amount of cash. Due to vacillating users demands and seasonal patterns, it is a very challenging problem for the financial institutions to keep the optimal amount of cash for each ATM. In this paper, we present a time series model based on Auto Regressive Integrated Moving Average (ARIMA) technique called Time Series ARIMA Model for ATM (TASM4ATM). This study used ATM back-end refilling historical data from 6 different financial organizations in Pakistan. There are 2040 distinct ATMs and 18 month of replenishment data from these ATMs are used to train the proposed model. The model is compared with the state-of- the-art models like Recurrent Neural Network (RNN) and Amazon’s DeepAR model. Two approaches are used for forecasting (i) Single ATM and (ii) clusters of ATMs (In which ATMs are clustered with similar cash-demands). The Mean Absolute Percentage Error (MAPE) and Symmetric Mean Absolute Percentage Error (SMAPE) are used to evaluate the models. The suggested model produces far better forecasting as compared to the models in comparison and produced an average of 7.86/7.99 values for MAPE/SMAPE errors on individual ATMs and average of 6.57/6.64 values for MAPE/SMAPE errors on clusters of ATMs.
... On looking at the literature, we find very little knowledge about the approach to resolve ATM cash management problems. Ekinci et al. proposed that ATMs be clustered and aimed to maximize the sum of routine cash withdrawals using forecasting techniques [12]. Their research found that these combined forecasting and optimization mechanisms worked well to reduce loading cash prices, consumer frustration, and cash interest rates. ...
Automated Teller Machine (ATM) service providers are constantly required to maintain a minimum amount of cash flow to effectively serve their customers. A successful balance between the need to provide ample money in ATMs and minimize cash interest costs sometimes leads to overstock or out-of-stock situations. Higher models of accuracy are required for thorough analysis and learning from past transactional data to predict an optimal requirement. In this research, the ATMs incorporate an integrated cash demand forecasting and stock optimization model to enhance the supply chain of a bank. We first conduct a thorough exploratory analysis to determine hidden patterns and apply a bidirectional long short-term memory (BLSTM) network with self-attention to improve the accuracy of the problem. The results proved that the applied model drastically reduced the rate of idle cash and prevented out-of-stock situations for all the ATM branches. We indicate that the proposed approach is highly effective by using the transactional data from an international commercial bank. The study, to the best of our knowledge, is the first to implement a novel approach using BLSTM with multi-head attention in resolving the ATM cash optimization problem, surpassing the state-of-the-art results.
... This cluster-level replenishment plans could result in saving huge operational costs for ATMs operating in a similar geographical region. Other papers of the existing literature focused on ATMs are (Van der Heide et al., 2020), (Ekinci et al., 2015), (Jadwal et al., 2018) or (Teddy and Ng, 2011) who optimize the ATM cash replenishment or develop different systems for predicting the daily amounts withdrawn from ATM´s. Under the Inventory management view, in (Nodoust et al., 2014) an approach for an inventory system is developed. ...
Some widely-accepted practices on banking ATM networks may be negatively affecting an efficient liquidity management. This paper analyses ATM cash management in light of empirical evidence which suggests that banking ATMs tend to be overloaded beyond the customer’s needs. This, in turn, results in high opportunity costs. While this is not perceived by banks as particularly harmful, it might have a damaging impact on other business which revolve exclusively around ATM networks, such as cashback sites.
Dormant money may be overcome by an appropriate tool matching the ATM’s cash to the user’s needs. Supported by a large database of banking records, this paper also provides model validation for a set of theorems previously developed by the author, resulting here in a cutting-edge, reliable forecasting system, suitable for anticipating ATMs cash demand as well as coupling with other supply chain planning processes.
... In addition to the IRP, demand forecasting for ATMs has been covered in, for instance, Van Anholt (2014) and Vanketesh et al. (2014). Moreover, the combination of these two aspects has been covered in Ekinci et al. (2015), where time intervals between two replenishments are determined using a forecast dependent on clusters of nearby ATMs. Furthermore, inventory management in the case of lost-sales has been covered extensively (Bijvank and Vis, 2011). ...
In the inventory management of automated teller machines (ATMs) many activities affect the total costs, such as forecasting, replenishments and the denomination mix used. The denomination mix is the combination of bills used to fulfill a customer’s demand. We investigate whether allowing the denomination mix to vary over time based on the forecast withdrawals at an ATM reduces actual operating costs of an ATM. To verify this, we propose a time-varying denomination mix strategy, which is validated by benchmarking it against the case of a bank’s denomination mix strategy. The bank’s predetermined strategy typically consists of a least note strategy or a one-smallest strategy. In all strategies we simultaneously optimize denomination mixes and replenishment decisions. We define the problem and solution strategies as mixed integer programming formulations and solve them via a rolling horizon algorithm using different frequencies of denomination mix updates, rolling horizon lengths, numbers of ATMs, cost parameters, and forecast qualities. By implementing the time-varying denomination mix, we show that the operational costs of managing an ATM can be reduced by 21% or €153.77 per ATM per month on average, which can represent over €10 million per year in the Netherlands only.
... Since the bank replenishes the ATMs weekly, we decided to take the time period as 1 week. Moreover, in our other study, (Ekinci et al. 2015), the results obtained from several experiments show that the forecasting model built from weekly data does a better job than models built from daily data. ...
The use of Automated Teller Machines (ATMs) has become increasingly popular throughout the world due to the widespread adoption of electronic financial transactions and better access to financial services in many countries. As the network of ATMs is becoming denser while the users are accessing them at a greater rate, the current financial institutions are faced with addressing inventory and replenishment optimal policies when managing a large number of ATMs. An excessive ATM replenishment will result in a large holding cost whereas an inadequate cash inventory will increase the frequency of the replenishments and the probability of stock-outs along with customer dissatisfaction. To facilitate informed decisions in ATM cash management, in this paper, we introduce an approach for optimal replenishment amounts to minimize the total costs of money holding and customer dissatisfaction by taking the replenishment costs into account including stock-outs. An important aspect of the replenishment strategy is that the future cash demands are not available at the time of planning. To account for uncertainties in unobserved future cash demands, we use prediction intervals instead of point predictions and solve the cash replenishment-planning problem using robust optimization with linear programming. We illustrate the application of the optimal ATM replenishment policy under future demand uncertainties using data consisting of daily cash withdrawals of 98 ATMs of a bank in Istanbul. We find that the optimization approach introduced in this paper results in significant reductions in costs as compared to common practice strategies.
... Related researches mostly studied cash demand prediction and loading time schedules separately (Ekinci, Lu, & Duman, 2015). In this paper, it is going to be shown that daily cash withdrawals are predictable for individual ATMs. ...
ATMs are physical interaction points between financial institutions and real customers. Storing physical cash causes renouncing to get interested. On the other hand, customer satisfaction requires to store the necessary cash amount. This concern becomes even more critical for countries having high-interest rate and overnight interest rates are higher. In this paper, we will show that daily cash withdrawals are predictable and we will propose a cost function for replenishment optimization. Experiments show that proposed model decrease idle balance dramatically.
... Ekinci, et al, suggested grouping of ATMs into nearbylocation clusters and also try to optimize the sum of daily cash withdraws with the forecasting techniques. Their studies showed that this integrated forecasting and optimization procedure performs well in order to minimize costs of loading cash, customer dissatisfaction, and cash-interest fee [16]. ...
... For instance, the study in [18] uses neural networks and least square support vector machines, while the work in [19] also uses artificial neural networks and neuro-fuzzy models and the one in [20] uses a local linear wavelet neural network for time-series prediction. While most works in the literature studied forecasting and ATM cash replenishment policy separately, the work in [21] proposes an integrated approach where the ATMs in near-by locations are grouped into clusters and the optimal replenishment time interval as well as data forecasting are applied to the clusters. Another work in [22], elaborates on a model based on the combination of neural networks and multiagent technology for predicting future cash demand. ...
Cash-related costs constitute a large portion of management cost of an automated teller machine (ATM) network. Cash should be delivered to or picked-up from ATM devices in certain intervals in order to both meet customer satisfaction and to be able to generate additional revenue from excess cash through daily interest rates. Unlike classical ATMs, new-generation ATMs, also called recycle ATMs, have a single cassette for cash withdrawal and deposit; this property imposes new restrictions on ATM cash management. Moreover, recycle ATMs are costly, and hence their deployment should be planned carefully. In this paper, our aim is to optimize the ATM networks in terms of cash related costs. We formulate an optimization problem as an integer linear program, which jointly decides on when to visit an ATM, how much money to deliver to which ATM and which road should be followed for the distribution of cash to the ATMs. We also decide on which ATMs in the network should be replaced by a recycle ATM. We then propose a polynomial-time heuristic algorithm and compare it with the optimization formulation in terms of cash cost and the recycle ATM decision. We demonstrate through performance evaluation that our heuristic algorithm is suitable for practical implementation.
... Many large companies try to find ways to utilise available historical data in order to draw forecasts for making decisions, thereby seeking to improve their manufacturing, logistics, or marketing operations (Ekinci et al., 2015). For example, Samsung uses its retail stores as a source of various market data, such as customers' sales quantity, inventory level, orders, and sales forecasts. ...
A common phenomenon that decreases the accuracy of time series
forecasting is the existence of change points in the data. This paper presents a
method for time series forecasting with the possibility of a change point in the
distribution of observations. The proposed method uses change point
techniques to detect and estimate change points, and to improve the forecasting
process by taking change points into account. The method can be applied to
both stationary series and linear trend series. Change point analysis prevents the
omission of relevant data as well as the forecasting that may be based on
irrelevant data. The study concludes that change point techniques may increase
the accuracy of forecasts, as is demonstrated in the real case study presented in
this paper.
... User can enter the required amount and the application finds out whether which ATM has so much balance in ATM and can suggest that ATM [8] [10]. ...
... Generally, RFID allows better identification with lower error rate of control process. RFID also reduces errors of ATM systems (Ekinci et al., 2015). ...
Identifying and tracking goods in the supply chain is one of the most important factors influencing supply chain management. This paper is trying to determine the effects of applying radio frequency identification (RFID) to identify items in the supply chain of manufacturing enterprises. To this aim, in this paper, at first, a comprehensive literature review and experts' judgments are employed to identify possible influential factors, and then a questionnaire is conducted for evaluating these factors. This questionnaire filled by about 200 experts of using RFID in manufacturing enterprise. Filled questionnaire used for examination of statistical hypothesis, which examine the impact of this technology on supply chain. The results showed that applying RFID has a positive impact on organisational performance, especially in supply chain management processes and inventory management. Also it can lead to reduce errors, improve performance, save time, increase security in payments and reduce theft, improve the quality of management decisions, increase job satisfaction and improve effective operational automation.
... Dawande et al. (2006) studied supply chain scheduling for distribution systems. Ekinci et al. (2015) proposed optimization of ATM cash replenishment with group-demand forecasts. This article proposed grouping ATMs into nearby-location clusters and also optimizing the aggregates of daily cash withdraws in the forecasting process. ...
Optimization issues of scheduling problems have long been examined for more than 50 years. Numerous studies are devoted to the scheduling issues in logistics and supply chain systems. Indeed, scheduling problems of logistics and supply chain systems have a combinatorial nature and thus computational complexity is usually very high. In this regard, this study reviews the existing literature and then examines an optimal scheduling problem of distributions to demanding nodes based on preferences. The scheduling problem is approached as a mixed integer program and an optimal solution is retrieved by using a solver methodology.
... Treating intraday economic time series as random continuous (transaction level) functions projected onto low dimensional subspace is another promising approach presented by Laukaitis (2008) who also proposed functional autoregressive model as a robust predictor of the cash flow and intensity of transactions in a credit card payment systems. Integrated forecasting, when ATM grouping into nearby-location clusters is performed with optimization of aggregates of daily cash withdrawals (Ekinci, Lu & Duman, 2015) can be effective alternative to individual cash flow time series forecasting. ...
... Researcher in [11] treats intraday cash flow time series as random continuous functions projected onto low dimensional subspace and use functional autoregressive model as predictor of cash flow and intensity of transactions. ATM clustering approach was employed by [12] when integrated forecasting of aggregation of nearby-location ATM cash demand was performed. ...
Good ATM network cash management requires accurate information of future cash demand. In this paper we compare computational intelligence models when performing cash flow forecasting for one day. Adaptive input selection and model parameter identification are used with every forecasting model in order to perform more flexible comparison. Experimental data contains 200 ATMs from real ATM network with historical period of 26 months. Investigation of historical data length influence for forecasting accuracy with every model is also performed. Results suggest-SVR (support vector regression) forecasting model performs best when SMAPE forecasting accuracy measure is used.
In the digital transformation era, integrating digital technology into every aspect of banking operations improves process automation, cost efficiency, and service level improvement. Although logistics for Automated Teller Machine (ATM) cash is a crucial task that impacts operating costs and consumer satisfaction, there has been little effort to enhance it. Specifically, in Vietnam, with a market of more than 20,000 ATMs nationally, research and technological solutions that can resolve this issue remain scarce. In this paper, we generalized the vehicle routing problem for ATM cash replenishment, suggested a mathematical model, and then offered a tool to evaluate various situations. When being evaluated on the simulated dataset, our proposed model and method produced encouraging results with the benefits of cutting ATM cash operating costs.
Logistics has emerged as a crucial component in various business domains, playing a significant role in ensuring efficient operations. In addition to traditional applications, logistics principles are also being applied in the financial sector, specifically in the management of Automated Teller Machines (ATMs). ATMs offer a self-service and time-independent mechanism, providing financial institutions with an efficient means of serving their customers. However, the network design of cash distribution poses several challenges that necessitate an optimized solution. This solution aims to fulfill customer demands for ATMs while simultaneously minimizing losses for banks. This paper proposes a combined approach to address these challenges, integrating the demand forecasting with the vehicle routing problem. The replenishment policy begins with forecasting cash withdrawals, utilizing various methods such as statistical methodologies (e.g., ARIMA and SARIMA) and machine learning techniques (e.g., Prophet and DNN). To determine optimal routes for armored trucks and minimize costs based on the forecasted data, the VRP Spreadsheet Solver tool is implemented. By developing a decision support system, several methods are applied to facilitate ATM visitation using inventory control methodologies and vehicle routing techniques. This integrated approach seeks to achieve a balance between meeting ATM customer demands and optimizing the utilization of resources in cash replenishment and distribution. Overall, this research presents a comprehensive solution for addressing the challenges in cash network design for ATMs. By combining forecasting methods with vehicle routing optimization, it offers a decision support system that enhances the efficiency of ATM operations while minimizing costs and ensuring customer satisfaction.KeywordsInventory ManagementForecastingVehicle RoutingDecision Support SystemATM
in collaboration with Robert Gordon University Aberdeen , UK 2021 ii Declaration I declare that the work presented in this dissertation is my own work and to best of my knowledge acknowledgement is made for all sources of information used in this dissertation. Further, this as a whole or as parts has not been submitted previously or concurrently for a degree or any other qualification at any University or institutions of Higher Learning. …………………………. ……………………. Signature of the student Date Full Name of Student: Student Registration No: The above student carried out his/her research project under my supervision. ………………………… ……………………. Signature of the supervisor Date Name of the Supervisor: iii Abstract Predicting sales has become a vital role in consumer durable retail industry with the increased competition. The importance of this has increased due to its direct and indirect affect to the company's profitability. Managing the right inventory would reduce inventory holding costs while overstocking brings repercussions such as increasing the inventory holding cost, occurring promotional costs to flush out the excess stock etc. In this study, several machine learning techniques were used to predict the sales using the factors affecting sales. The techniques namely; K-nearest neighbor (KNN), Artificial neural network (ANN), Decision tree, Random Forest, Linear Regression and Bayesian Applied Regression. This study was conducted on four main product categories namely; Air conditioners, Fans, Refrigerator and audios. The algorithms were applied on the preprocessed sales data captured from the ERP system of the company, promotional data, event data and weather data. Once applied the algorithms, the best model was selected comparing RMSE and MAPE. KNN was identified to be the best fitting model of the seven models used for all the product lines. The analysis were done using R programming language and then the predicted values are then visualized through Power BI dashboards to present to the management. These predictions visualized through dashboards allows the retailer to take correct decisions at the right time while monitoring the progress of the actions. Moreover, the finalized data model allows the retailer to gain the cutting edge over the competitors while increasing the profits. iv
The main function of the internal control department of a bank is to inspect the banking operations to see if they are performed in accordance with the regulations and bank policies. To accomplish this, they pick up a number of operations that are selected randomly or by some rule and, inspect those operations according to some predetermined check lists. If they find any discrepancies where the number of such discrepancies are in the magnitude of several hundreds, they inform the corresponding department (usually bank branches) and ask them for a correction (if it can be done) or an explanation. In this study, we take up a real-life project carried out under our supervisory where the aim was to develop a set of predictive models that would highlight which operations of the credit department are more likely to bear some problems. This multi-classification problem was very challenging since the number of classes were enormous and some class values were observed only a few times. After providing a detailed description of the problem we attacked, we describe the detailed discussions which in the end made us to develop six different models. For the modeling, we used the logistic regression algorithm as it was preferred by our partner bank. We show that these models have Gini values of 51 per cent on the average which is quite satisfactory as compared to sector practices. We also show that the average lift of the models is 3.32 if the inspectors were to inspect as many credits as the number of actual problematic credits.
The use of the social media (SM) has become more and more widespread during the last two decades, the companies started looking for insights for how they can improve their businesses using the information accumulating therein. In this regard, it is possible to distinguish between two lines of research: those based on anonymous data and those based on customer specific data. Although obtaining customer specific SM data is a challenging task, analysis of such individual data can result in very useful insights. In this study we take up this path for the customers of a bank, analyze their tweets and develop three kinds of analytical models: clustering, sentiment analysis and product propensity. For the latter one, we also develop a version where, besides the text information, the structural information available in the bank databases are also used in the models. The result of the study is a considerably more efficient set of analytical CRM models.
ATM cash management is a set of transactions to optimize the amount of money that should be kept on ATM devices. ATM Cash Management Optimization studies are the studies carried out to keep the most appropriate amount of cash on the ATM by the types and models of ATMs in the banking system, following each time zone. In this study, ATM cash management will be handled as an optimization problem, and the decision variable values that will minimize the treasury and operation costs will be calculated. The main purpose of the study is to estimate the most appropriate amount of money that should be in ATM devices daily and to recommend the amount of cash that should be in the ATM to the system owners. In this study, the most ideal result was tried to be found by using metaheuristic multi objective optimization algorithms used in the literature. First of all, the decision variables and features of ATM data were determined by a detailed study. The optimization algorithm was decided to produce results with a new approach for the determined data areas. The arithmetic Optimization Algorithm (AOA) algorithm, one of the newly proposed metaheuristic algorithms, was chosen as the optimization algorithm. The fuzzy logic algorithm was used to find the weighting coefficients of the values used as input parameters for the AOA algorithm. By using the fuzzy logic algorithm, the effect values of the features on the result were weighted, and together with the obtained weighting information, it was proposed as research that the most appropriate cash amount value should be used in the AOA algorithm. In this way, it is aimed to present a successful solution by exhibiting a hybrid approach. The results obtained were compared with the values in the system and the success rates were tried to be revealed.KeywordsFuzzy logicAOAArithmetic Optimization AlgorithmCash ManagementOptimizationATM
One of the rare departments that data scientists have not deal with yet is the Internal Audit department where they inspect and try to find out the bank personnel who might have been committing fraud. The main reason of this is the too few known fraudulent cases and thus an extremely imbalanced data between fraudulent and legitimate transactions. On the other hand, the work of the auditors is too complicated as they touch almost all banking system.
In this study, we, for the first time in the literature, apply data analytics tools in a novel framework to help the auditors. The aim is to indicate which transactions, or personnel should the auditors inspect in the first place so that they can figure out problems with a higher chance. To overcome the difficulties related with the extreme imbalance of the data, we suggested and compared two different ideas. First one is making use of descriptive techniques together with the predictive tools. The second one is using a genuine technique to oversample the minority class. The models developed are shown to perform good on past data.
Purpose
The research objective is to increase the computational efficiency of the automated teller machine (ATM) cash demand forecasting problem. It proposes a practical decision-making process that uses aggregated time series of a bank's ATM network. The purpose is to decrease ATM numbers that will be forecasted by individual models, by finding the machines’ cluster where the forecasting results of the aggregated series are appropriate to use.
Design/methodology/approach
A comparative statistical forecasting approach is proposed in order to reduce the calculation complexity of an ATM network by using the NN5 competition data set. Integrated autoregressive moving average (ARIMA) and its seasonal version SARIMA are fitted to each time series. Then, averaged time series are introduced to simplify the forecasting process carried out for each ATM. The ATMs that are forecastable with the averaged series are identified by calculating the forecasting accuracy change in each machine.
Findings
The proposed approach is evaluated by different error metrics and is compared to the literature findings. The results show that the ATMs that have tolerable accuracy loss may be considered as a cluster and can be forecasted with a single model based on the aggregated series.
Research limitations/implications
The research is based on the public data set. Financial institutions do not prefer to share their ATM transactions data, therefore accessible data are limited.
Practical implications
The proposed practical approach will be beneficial for financial institutions to use, that hold an excessive number of ATMs because it reduces the computational time and resources allocated for the forecasting process.
Originality/value
This study offers an effective simplified methodology to the challenging cash demand forecasting process by introducing an aggregated time series approach.
In today’s competitive banking industry, deciding on the launch of new facilities is very important, and the lack of attention to this issue will lead to heavy costs. Automated Teller Machine (ATM) Branch is one of the most important facilities in the banking industry. In the current competitive market, choosing the optimal location for these facilities beside an optimal weekly routing for their cash replenishment is very important from the economic and environmental aspects. In this study, a novel mathematical model is presented to integrate the two well-studied location and vehicle routing problems for ATMs in banking industry. In location side, considering the cost of deployment, the model tries to find the optimal new ATMs location to maximize the coverage of ATM branches and the bank’s share in competing with branches of other existing banks by using the different elements of a gravity function. Then, by aware of the location of the new facilities, the model concurrently attempts to provide an optimal cash replenishment policy by embedding a green vehicle routing problem and taking into account a central warehouse and several types of banknote, in order to minimize total costs including the transportation, disposition and shortage costs such that the total GHG emissions generated by the cash carrier vehicles are also minimized. The proposed model is applied in a real case study and the results show that the model can be effectively used by the bankers to increase the performance of the ATMs network and pragmatically contribute to social efforts in response to the environmental concerns.
With the automated teller machine (ATM) cash replenishment problem, banks aim to reduce the number of out-of-cash ATMs and duration of out-of-cash status. On the other hand, they want to reduce the cost of cash replenishment, as well. The problem conventionally involves forecasting ATM cash withdrawals, and then cash replenishment optimization based on the forecast. The authors assume that reliable forecasts are already obtained for the amount of cash needed in ATMs. The focus of the article is cash replenishment optimization. After introducing linear programming-based solutions, the authors propose a solution based on dynamic programming. Experiments conducted on real data reveal that the proposed approach can find the optimal solution more efficiently than linear programming.
The main focus in atm cash replenishment, helps to maintain the operational cost in banking sector. It advocates the grouping atm into clusters with corresponding withdrawl patterns. LSTM method has used to recuperate the cash to all atm.
This paper describes how Machine Learning and Robust Optimization techniques can greatly improve cash logistics operations. Specifically, we seek to optimize the logistics followed by the different branches of a given bank. Machine Learning is used to forecast cash demands for each of the branches, taking into account past demands and calendar effects. These demand predictions are forwarded to a Robust Optimization model, whose outputs are the cash transports that each branch should request. These transports guarantee that demand is fulfilled up to the desired confidence level, while also satisfying additional constraints arising in this particular domain.
Forecasting cash management, security, ease of use, and so on are important in the use of Automated Teller Machine (ATM). For this purpose, in this paper, we have discussed issues such as forecasting cash demand, fraud detection, ATM failure, user interface, replenishment strategy, ATM location, customer behavior, etc. Artificial Intelligence (AI) techniques are discussed for the detection of fraud, failure, replenishment and crash prediction. A number of statistical methods used to evaluate these forecasts are also covered in this paper. Moreover, we review AI techniques such as neural networks, regressions, support vector machines and their results in the form graphs in different sections. The literature covered in this paper is related to the past ten years (2006-2016). The approaches studied in this paper are compared in terms of data sets and prediction performance, accuracy and so on.We also provide a list of data sets available for the scientific community to conduct research in this field. Finally, open issues and future works in each of these items are presented.
The purpose of cash management is to optimize distribution of cash. Effective cash management brings savings to retail banks that are related to: dormant cash reduction; reduced replenishment costs; decrease of cash preparation costs; reduction of cash insurance costs. Optimization of cash distribution for retail banking in ATM and branch networks requires estimation of cash demand/supply in the future. This estimation determines overall cash management efficiency: accurate cash demand estimation reduces bank overall costs. In order to estimate cash demand in the future, cash flow forecasting must be performed that is usually based on historical cash point (ATM or branch) cash flow data. Many factors that are uncertain and may change in time influence cash supply/demand process for cash point. These may change throughout cash points and are related to location, climate, holiday, celebration day and special event (such as salary days and sale of nearby supermarket) factors. Some factors affect cash demand periodically. Periodical factors form various seasonality in cash flow process: daily (related to intraday factors throughout the day), weekly (mostly related to weekend effects), monthly (related to payday) and yearly (related to climate seasons, tourist and student arrivals, periodical celebration days such as New Year) seasons. Uncertain (aperiodic) factors are mostly related to celebration days that do not occur periodically (such as Easter), structural break factors that form long term or permanent cash flow shift (new shopping mall near cash point, shift of working hours) and some may be temporal (reconstruction of nearby building that restricts cash point reachability). Those factors form cash flow process that contains linear or nonlinear trend, mixtures of various seasonal components (intraday, weekly, monthly yearly), level shifts and heteroscedastic uncertainty. So historical data-based forecasting models need to be able to approximate historical cash demand process as accurately as possible properly evaluating these factors and perform forecasting of cash flow in the future based on estimated empirical relationship.
As part of its overall effort to maintain good customer service while managing operational efficiency and reducing cost, a bank in Singapore has embarked on using data and decision analytics methodologies to perform better ad-hoc ATM failure forecasting and plan the field service engineers to repair the machines. We propose using a combined Data and Decision Analytics Framework which helps the analyst to first understand the business problem by collecting, preparing, and exploring data to gain business insights, before proposing what objectives and solutions can and should be done to solve the problem. This paper reports the work in analyzing past daily ad-hoc ATM failures, forecasting ad-hoc ATM failures and then using the forecasted results to optimize the number of field service engineers to deploy in each geographical zone, to minimize the number of daily unattended ad-hoc ATM failures. The optimization model ensures that the least number of engineers are deployed in each zone on each day. However, to maintain a consistent number of engineers for a 2-week schedule, we recommend to deploy the maximum number of engineers in each zone within the 2 weeks. The resulting surplus engineer idle hours is reduced, and it represents a cost savings of 28.6% when compared with the bank's current practice.
Improving the ATM cash management techniques of banks has already received significant attention in the literature as a separate optimisation problem for banks and the independent firms that supply cash to automated teller machines. This article concentrates instead on a further possibility of cost reduction: optimising the cash management problem as one single problem. Doing so, contractual prices between banks and the cash in transit firms can be in general modified allowing for further cost reduction relative to individual optimisations. In order to show the pertinence of this procedure, we have determined possible Pareto-improvement re-contracting schemes based on a Baumol-type cash demand forecast for a Hungarian commercial bank resulting in substantial cost reduction.
The purpose of cash management is to optimize distribution of cash. Effective cash management brings savings to retail banks that are related to: dormant cash reduction; reduced replenishment costs; decrease of cash preparation costs; reduction of cash insurance costs.
Optimization of cash distribution for retail banking in ATM and branch networks requires estimation of cash demand/supply in the future. This estimation determines overall cash management efficiency: accurate cash demand estimation reduces bank overall costs. In order to estimate cash demand in the future, cash flow forecasting must be performed that is usually based on historical cash point (ATM or branch) cash flow data.
Many factors that are uncertain and may change in time influence cash supply/demand process for cash point. These may change throughout cash points and are related to location, climate, holiday, celebration day and special event (such as salary days and sale of nearby supermarket) factors. Some factors affect cash demand periodically. Periodical factors form various seasonality in cash flow process: daily (related to intraday factors throughout the day), weekly (mostly related to weekend effects), monthly (related to payday) and yearly (related to climate seasons, tourist and student arrivals, periodical celebration days such as New Year) seasons. Uncertain (aperiodic) factors are mostly related to celebration days that do not occur periodically (such as Easter), structural break factors that form long term or permanent cash flow shift (new shopping mall near cash point, shift of working hours) and some may be temporal (reconstruction of nearby building that restricts cash point reachability).
Those factors form cash flow process that contains linear or nonlinear trend, mixtures of various seasonal components (intraday, weekly, monthly yearly), level shifts and heteroscedastic uncertainty. So historical data-based forecasting models need to be able to approximate historical cash demand process as accurately as possible properly evaluating these factors and perform forecasting of cash flow in the future based on estimated empirical relationship.
The aim of this research is to study how cash flow process factors affect cash flow forecasting accuracy in ATM network, using computational intelligence methods as cash flow forecasting models when performing daily aggregated cash flow forecasting. For factor evaluation 8 typical (affected by different factors) ATM cash withdrawal process flows selected from real ATM network are used with historical period of 33 months.
The aim of this paper is to present the results of the time series forecasting competition that was organized within the IFSA-EUSFLAT 2015 conference.
This research studies a cash inventory problem in an ATM Network to satisfy customer's cash needs over multiple periods with deterministic demand. The objective is to determine the amount of money to place in Automated Teller Machines (ATMs) and cash centers for each period over a given time horizon. The algorithms are designed as a multi-echelon inventory problem with single-item capacitated lot-sizing to minimize total costs of running ATM network. In this study, we formulate the problem as a Mixed Integer Program (MIP) and develop an approach based on reformulating the model as a shortest path formulation for finding a near-optimal solution of the problem. This reformulation is the same as the traditional model, except the capacity constraints, inventory balance constraints and setup constraints related to the management of the money in ATMs are relaxed. This new formulation gives more variables and constraints, but has a much tighter linear relaxation than the original and is faster to solve for short term planning. Computational results show its effectiveness, especially for large sized problems.
The purpose of this paper is to introduce, model, and solve a rich multiperiod inventory-routing problem with pickups and deliveries motivated by the replenishment of automated teller machines in the Netherlands. Commodities can be brought to and from the depot, as well as being exchanged among customers to efficiently manage their inventory shortages and surpluses. A single customer can both provide and receive commodities at different periods, since its demand changes dynamically throughout the planning horizon and can be either positive or negative. In the case study, new technology provides these machines with the additional functionality of receiving deposits and reissuing banknotes to subsequent customers. We first formulate the problem as a very large-scale mixed-integer linear programming model. Given the size and complexity of the problem, we first decompose it into several more manageable subproblems by means of a clustering procedure, and we further simplify the subproblems by fixing some variables. The resulting subproblems are strengthened through the generation of valid inequalities and solved by branch and cut. We assess the performance of the proposed solution methodology through extensive computational experiments using real data. The results show that we are able to obtain good lower and upper bounds for this new and challenging practical problem.
The mass of the issues facing today's ATMs, cash management optimization can help banks to manage the system dynamically. The main question in this optimization is, forecasting demand for cash from the ATM, at specified time intervals (e.g., daily or weekly or monthly). Since the structure of the signal (the predicted demand for cash) is nonlinear with many factors, including: days, weeks, days, months, holidays, and even the location of the ATM is connected, the common methods of identification is usually not a good answer and should be used soft computing techniques such as fuzzy logic and neural networks. In this study, the trend methods used in this field explores the application of fuzzy logic (type II) in this issue has been dealt with.
This paper deals with the problem of withdrawals from Automated Teller Machines (ATMs), using daily data for selected ATMs installed by the Euronet network in the Polish provinces of Malopolska and Podkarpacie for the period from January 2008 to March 2012. The main aim of this paper is an estimation of the proper econometric models for withdrawals time series and attempt to forecast future demand on cash flow in ATMs in respect to their localization. This is necessary to establish a replenishment schedule. The results of computations suggest that models built on the basis of SARIMA methodology are useful tools for an modeling daily withdrawals time series. This kind of model can be applied independently of the localization of an ATM. The exercises for ex post data imply ex post forecast errors under 20%. This size of forecast errors is lower than the bias of actual replenishment scheduling.
ABSTRACTA number of market changes are impacting the way financial institutions are managing their automated teller machines (ATMs). We propose a new class of adaptive data‐driven policies for a stochastic inventory control problem faced by a large financial institution that manages cash at several ATMs. Senior management were concerned that their current cash supply system to manage ATMs was inefficient and outdated, and suspected that using improved cash management could reduce overall system cost. Our task was to provide a robust procedure to tackle the ATM's cash deployment strategies. Current industry practice uses a periodic review system with infrequent parameter updates for cash management based on the assumption that demand is normally distributed during the review period. This assumption did not hold during our investigation, warranting a new and robust analysis. Moreover, we discovered that forecast errors are often not normally distributed and that these error distributions change dramatically over time. Our approach finds the optimal time series forecaster and the best‐fitting weekly forecast error distribution. The guaranteed optimal target cash inventory level and time between orders could only be obtained through an optimization module that was embedded in a simulation routine that we built for the institution. We employed an exploratory case study methodology to collect cash withdrawal data at 21 ATMs owned and operated by the financial institution. Our new approach shows a 4.6% overall cost reduction. This reflects an annual cost savings of over $250,000 for the 2,500 ATM units that are operated by the bank.
The paper presents an artificial neural network based approach in support of cash demand forecasting for automatic teller machine (ATM). On the start phase a three layer feed-forward neural network was trained using Levenberg-Marquardt algorithm and historical data sets. Then ANN was retuned every week using the last observations from ATM. The generalization properties of the ANN were improved using regularization term which penalizes large values of the ANN weights. Regularization term was adapted online depending on complexity of relationship between input and output variables. Performed simulation and experimental tests have showed good forecasting capacities of ANN. At current stage the proposed procedure is in the implementing phase for cash management tasks in ATM network.
Determining an optimized amount of cash in bank’s Automatic Teller Machines (ATM) is a tricky job,
as the demand for cash fluctuates due to change in customer’s behavior, preferences, seasonality, time etc. The
decision of optimized cash refilling in ATM is done manually according to corporate policies and past
experience. This process may sometimes lead to poor service or unnecessary cost due to under or over-estimation
of cash demand. For this reason, finding the best match between cash requirement and demand fulfillment
becomes a crucial decision for bank authorities. Therefore, the purpose for banks is to decide an optimum
amount of money that should be placed in ATM to minimize opportunity costs and at the same time to satisfy
the customers’ untimely and uncertain requirements.
The paper suggests an application of fuzzy ARTMAP Network for proactively analyzing and forecasting daily
cash requirement in ATM assuring prompt cash availability and dispensing service. Parameter selection is
performed using neighborhood mutual information-based algorithm for attribute reduction to find best
parameters. Simulation results for ATM cash forecasting show the feasibility and effectiveness of the proposed
method.
To improve ATMs' cash demand forecasts, this paper advocates the prediction of cash demand for groups of ATMs with similar day-of-the week cash demand patterns. We first clustered ATM centers into ATM clusters having similar day-of-the week withdrawal patterns. To retrieve "day-of-the-week" withdrawal seasonality parameters (effect of a Monday, etc.) we built a time series model for each ATMs. For clustering, the succession of seven continuous daily withdrawal seasonality parameters of ATMs is discretized. Next, the similarity between the different ATMs' discretized daily withdrawal seasonality sequence is measured by the Sequence Alignment Method (SAM). For each cluster of ATMs, four neural networks viz., general regression neural network (GRNN), multi layer feed forward neural network (MLFF), group method of data handling (GMDH) and wavelet neural network (WNN) are built to predict an ATM center's cash demand. The proposed methodology is applied on the NN5 competition dataset. We observed that GRNN yielded the best result of 18.44% symmetric mean absolute percentage error (SMAPE), which is better than the result of Andrawis, Atiya, and El-Shishiny (2011). This is due to clustering followed by a forecasting phase. Further, the proposed approach yielded much smaller SMAPE values than the approach of direct prediction on the entire sample without clustering. From a managerial perspective, the clusterwise cash demand forecast helps the bank's top management to design similar cash replenishment plans for all the ATMs in the same cluster. This cluster-level replenishment plans could result in saving huge operational costs for ATMs operating in a similar geographical region. (c) 2013 Elsevier B.V. All rights reserved.
This paper addresses the problem of forecasting irregular demand, balancing the tradeoff between forecast accuracy and cost of collecting information. The literature suggests the adoption of a clustering approach, however it is not clear under which conditions this method is actually beneficial. We consider three kinds of demand variability, namely structural (e.g. seasonality), managerial (e.g. promotions) and random (i.e. unpredictable), and we investigate their impact on the correlation of demand within clusters of customers and thus on the clustering approach effectiveness. We develop an analytical model of this relationship and test it with real data in the fresh food industry. Results show that while structural and managerial variability make the clustering approach feasible, random variability works in the opposite direction, providing guidelines on when this forecasting method can be adopted.
The purpose of this paper is to develop a simple formula to predict the distance traveled by fleets of vehicles in physical distribution problems involving a depot and its area of influence. Since the transportation cost of operating a break-bulk terminal (or a warehouse) is intimately related to the distance traveled, the availability of such a simple formula should facilitate the study of more complex logistics problems. A simple manual dispatching strategy intended to mimic what dispatchers do, but simple enough to admit analytical modeling is presented. Since the formulas agree rather well with the length of (nearly optimal) computer built tours, the predictions should approximate distances achievable in practice; the formulas seem realistic.
Purpose
In many industrial contexts, firms are encountering increasingly uncertain demand. Numerous factors are driving this phenomenon; however, a major change that is spreading among different sectors is the ever‐growing attention to customers. Companies have identified that customers are critical not only because they directly influence the success of specific products or firms, but also because they play a fundamental role in many internal processes. Although the role of customers in business processes has been deeply analysed, the issue of demand forecasting and the role of customers has not been fully explored. The present study aims to examine the impact of heterogeneity of customer requests on demand forecasting approaches, based on three action research cases. Based on the analysis of customer behaviour, an appropriate methodology for each case is designed based on clustering customers according to their demand patterns.
Design/methodology/approach
Objectives are achieved by means of three action research case studies, developed in cooperation with three different companies. The paper structures a general methodology based on these three experiences to help managers in better dealing with uncertain demand.
Findings
By means of proper analysis of customers' heterogeneity and by using simple statistical techniques such as cluster analysis, forecasting performance can significantly improve. In these terms, this work claims that focusing on customers' heterogeneity is a relevant topic both for practitioners and researchers.
Originality/value
The paper proposes some specific guidelines to forecast demand where customers' differences impact significantly on demand variability. In these terms, results are relevant for practitioners. Moreover, the paper claims that this issue should be better analysed in future researches and proposes some guidelines for future works.
In a merchandise depth test, a retail chain introduces new products at a small sample of selected stores for a short period prior to the primary selling season and uses the observed sales to forecast demand for the entire chain. We describe a method for resolving two key questions in merchandise testing: (1) which stores to use for the test and (2) how to extrapolate from test sales to create a forecast of total season demand for each product for the chain. Our method uses sales history of products sold in a prior season, similar to those to be tested, to devise a testing program that would have been optimal if it had been applied to this historical sample. Optimality is defined as minimizing the cost of conducting the test, plus the cost of over- and understocking of the products whose supply is to be guided by the test.
To determine the best set of test stores, we apply a k-median model to cluster the stores of the chain based on a store similarity measure defined by sales history, and then choose one test store from each cluster. A linear programming model is used to fit a formula that is then used to predict total sales from test sales.
We applied our method at a large retailer that specializes in women's apparel and at two major shoe retailers, comparing results in each case to the existing process used by the apparel retailer and to some standard statistical approaches such as forward selection and backward elimination. We also tested a version of our method in which clustering was based on a combination of several store descriptors such as location, type of store, ethnicity of the neighborhood of location, total store sales, and average temperature of the store location. We found that relative to these other methods, our approach could significantly improve forecasts and reduce markdowns that result from excessive inventory, and lost margins resulting from stockouts. At the apparel retailer the improvement was enough to increase profits by more than 100%.
We believe that one reason our method outperforms the forward selection and backward elimination methods is that these methods seek to minimize squared errors, while our method optimizes the true cost of forecast errors. In addition, our approach, which is based purely on sales, outperforms descriptor variables because it is not always clear which are the best store descriptors and how best to combine them. However, the sales-based process is completely objective and directly corresponds to the retailer's objective of minimizing the understock and overstock costs of forecast error.
We examined the stores within each of the clusters formed by our method to identify common factors that might explain their similar sales patterns. The main factor was the similarity in climate within a cluster. This was followed by the ethnicity of the neighborhood where the store is located, and the type of store. We also found that, contrary to popular belief, store size and location had little impact on sales patterns.
In addition, this technique could also be used to determine the inventory allocation to individual stores within a cluster and to minimize lost demand resulting from inaccurate distribution across size. Finally, our method provides a logical framework for implementing micromerchandising, a practice followed by a significant number of retailers in which a unique assortment of merchandise is offered in each store (or a group of similar stores) tuned to maximize the appeal to customers of that store. Each cluster formed by our algorithm could be treated as a "virtual chain" within the larger chain, which is managed separately and in a consistent manner in terms of product mix, timing of delivery, advertising message, and store layout.
Forecasting cash demands at automatic teller machines (ATMs) is challenging, due to the heteroskedastic nature of such time series. Conventional global learning computational intelligence (CI) models, with their generalized learning behaviors, may not capture the complex dynamics and time-varying characteristics of such real-life time series data efficiently. In this paper, we propose to use a novel local learning model of the pseudo self-evolving cerebellar model articulation controller (PSECMAC) associative memory network to produce accurate forecasts of ATM cash demands. As a computational model of the human cerebellum, our model can incorporate local learning to effectively model the complex dynamics of heteroskedastic time series. We evaluated the forecasting performance of our PSECMAC model against the performances of current established CI and regression models using the NN5 competition dataset of 111 empirical daily ATM cash withdrawal series. The evaluation results show that the forecasting capability of our PSECMAC model exceeds that of the benchmark local and global-learning based models.
We propose a simple way of predicting time series with recurring seasonal periods. Missing values of the time series are estimated and interpolated in a preprocessing step. We combine several forecasting methods by taking the weighted mean of forecasts that were generated with time-domain models which were validated on left-out parts of the time series. The hybrid model is a combination of a neural network ensemble, an ensemble of nearest trajectory models and a model for the 7-day cycle. We apply this approach to the NN5 time series competition data set.
In this work the self-organizing fuzzy neural network (SOFNN) is employed to create an accurate and easily calibrated approach to multiple-step-ahead prediction for the NN5 forecasting competition 2008. The competition dataset consists of 111 daily empirical time series of cash-machine withdrawals. The objective for the competition was to forecast future transactions up to 56 days ahead with the highest prediction accuracy using a single methodology. The SOFNN is a highly efficient and accurate algorithm for time series-prediction which learns from data incrementally and can autonomously adapt its structure in the learning process to cope with drifts in the data dynamics. It can also modify its architecture autonomously to suit different prediction horizons, embedding dimensions and time lags. Standard neural networks(NNs) and autoregressive(AR) models are employed as benchmarks for comparison. It is shown through a statistical analysis of the results, that the SOFNN significantly outperforms the NN and AR methods.
This paper considers the problem of constructing order batches for distribution centers using a data mining technique. With the advent of supply chain management, distribution centers fulfill a strategic role of achieving the logistics objectives of shorter cycle times, lower inventories, lower costs and better customer service. Many companies consider both their cost effectiveness and market proficiency to depend primarily on efficient logistics management. Warehouse management system (WMS) presently is considered a key to strengthening company logistics. Order picking is routine in distribution centers. Before picking a large set of orders, effectively grouping orders into batches can accelerate product movement within the storage zone. The order batching procedure has to be implemented in WMS and may be run online many times daily. The literature has proposed numerous batching heuristics for minimizing travel distance or travel time. This paper presents a clustering procedure for an order batching problem in a distribution center with a parallel-aisle layout. A data mining technique of association rule mining is adopted to develop the order clustering approach. Performance comparisons between the developed approach and existing heuristics are given for various problems.
The development of a system for predicting the daily amounts withdrawn from automated teller machines (ATMs) for inventory control is considered, using data from 190 ATMs in the United Kingdom over a two-year period. We argue that density forecasts are more appropriate than point forecasts and that a good forecasting system might choose a different model for each ATM. An analysis of the data finds that seasonal structure, first-order autocorrelation and cash-out days are important aspects of the data. Predictive sequential (prequential) comparisons between linear models, autoregressive models, structural time series models and Markov-switching models are made. The Markov-switching models are preferred because they are found to produce better density forecasts, and might also be more useful for inventory control because they separate the demand for cash from ‘out-of-service’ effects. A logarithmic scoring rule is used to choose the most appropriate seasonal and distributional assumptions for each ATM.
In this work we introduce the forecasting model with which we participated in the NN5 forecasting competition (the forecasting of 111 time series representing daily cash withdrawal amounts at ATM machines). The main idea of this model is to utilize the concept of forecast combination, which has proven to be an effective methodology in the forecasting literature. In the proposed system we attempted to follow a principled approach, and make use of some of the guidelines and concepts that are known in the forecasting literature to lead to superior performance. For example, we considered various previous comparison studies and time series competitions as guidance in determining which individual forecasting models to test (for possible inclusion in the forecast combination system). The final model ended up consisting of neural networks, Gaussian process regression, and linear models, combined by simple average. We also paid extra attention to the seasonality aspect, decomposing the seasonality into weekly (which is the strongest one), day of the month, and month of the year seasonality.
The treasurer of a bank is responsible for the cash management of several banking activities. In this work, we focus on two of them: cash management in automatic teller machines (ATMs), and in the compensation of credit card transactions. In both cases a decision must be taken according to a future customers demand, which is uncertain. From historical data we can obtain a discrete probability distribution of this demand, which allows the application of stochastic programming techniques. We present stochastic programming models for each problem. Two short-term and one mid-term models are presented for ATMs. The short-term model with fixed costs results in an integer problem which is solved by a fast (i.e. linear running time) algorithm. The short-term model with fixed and staircase costs is solved through its MILP equivalent deterministic formulation. The mid-term model with fixed and staircase costs gives rise to a multi-stage stochastic problem, which is also solved by its MILP deterministic equivalent. The model for compensation of credit card transactions results in a closed form solution. The optimal solutions of those models are the best decisions to be taken by the bank, and provide the basis for a decision support system.
Selecting optimal locations for new facilities is a critical decision in organizations that provide field-based services such as delivery, maintenance and emergency services. The total logistics cost and facility establishment cost are the main objectives of the location selection procedure. With the increasing size of this problem in today's applications, the aspects of efficiency and scalability have developed into major challenges. In this paper, we study the use of spatial clustering methods to solve this problem and propose two new algorithms. The new algorithms determine the optimal locations of the new facilities plus their optimal total count during the search process. We have conducted many experiments for empirical comparative study on the application of several spatial clustering algorithms for optimal facility establishment. The benchmarks are conducted with both real world and synthetic data sets. The results reveal advantages of the proposed algorithms and confirm that these algorithms have better performance in terms of efficiency and objectives in the field-based services. Hence, the higher scalability and effectiveness of the proposed algorithms make them suitable solutions for the problem of optimal facility establishment with large databases.
Albert Heijn, BV, a supermarket chain in the Netherlands, faces a vehicle routing and delivery scheduling problem once every three to six months. Given hourly demand forecasts for each store, travel times and distances, cost parameters, and various transportation constraints, the firm seeks to determine a weekly delivery schedule specifying the times when each store should be replenished from a central distribution center, and to determine the vehicle routes that service these requirements at minimum cost. We describe the development and implementation of a system to solve this problem at Albert Heijn. The system resulted in savings of 4% of distribution costs in its first year of implementation and is expected to yield 12%-20% savings as the firm expands its usage. It also has tactical and strategic advantages for the firm, such as in assessing the cost impact of various logistics and marketing decisions, in performance measurement, and in competing effectively through reduced lead time and increased frequency of replenishment.
Success in forecasting and analyzing sales for given goods or services can mean the difference between profit and loss for an accounting period and, ultimately, the success or failure of the business itself. Therefore, reliable prediction of sales becomes a very important task. This article presents a novel sales forecasting approach by the integration of genetic fuzzy systems (GFS) and data clustering to construct a sales forecasting expert system. At first, all records of data are categorized into k clusters by using the K-means model. Then, all clusters will be fed into independent GFS models with the ability of rule base extraction and data base tuning. In order to evaluate our K-means genetic fuzzy system (KGFS) we apply it on a printed circuit board (PCB) sales forecasting problem which has been used as the case in different studies. We compare the performance of an extracted expert system with previous sales forecasting methods using mean absolute percentage error (MAPE) and root mean square error (RMSE). Experimental results show that the proposed approach outperforms the other previous approaches.
Multi-step ahead forecasting is still an open challenge in time series
forecasting. Several approaches that deal with this complex problem have been
proposed in the literature but an extensive comparison on a large number of
tasks is still missing. This paper aims to fill this gap by reviewing existing
strategies for multi-step ahead forecasting and comparing them in theoretical
and practical terms. To attain such an objective, we performed a large scale
comparison of these different strategies using a large experimental benchmark
(namely the 111 series from the NN5 forecasting competition). In addition, we
considered the effects of deseasonalization, input variable selection, and
forecast combination on these strategies and on multi-step ahead forecasting at
large. The following three findings appear to be consistently supported by the
experimental results: Multiple-Output strategies are the best performing
approaches, deseasonalization leads to uniformly improved forecast accuracy,
and input selection is more effective when performed in conjunction with
deseasonalization.
Inventory routing problems combine the optimization of product deliveries (or pickups) with inventory control at customer sites. Our application concerns the planning of single product pickups over time; each site accumulates stock at a deterministic rate; the stock is emptied on each visit. At the tactical planning stage considered here, our objective is to minimize a surrogate measure of routing cost while achieving some form of regional clustering by partitioning the sites between the vehicles. The fleet size is given but can potentially be reduced. Planning consists in assigning customers to vehicles in each time period, but the routing, i.e., the actual sequence in which vehicles visit customers, is considered as an ``operational'' decision. The planning is due to be repeated over the time horizon with constrained periodicity. We develop a truncated branch-and-price-and-cut algorithm combined with rounding and local search heuristics that yields both primal solutions and dual bounds. On a large scale test problem coming from industry, we obtain a solution within 6.25% deviation from the optimal. A rough comparison between an operational routing resulting from our tactical solution and the industrial practice shows a 10% decrease in number of vehicles as well as in travel distance. The key to the success of the approach is the use of a state-space relaxation technique in formulating the master program to avoid the symmetry in time.
Cash inventory management at automated teller machines under incomplete information (Master thesis): The Department of Industrial Engıneering and The
- Y Altunoglu
Altunoglu, Y. (2010). Cash inventory management at automated teller machines under incomplete information (Master thesis): The Department of Industrial Engıneering and The Institute of Engineering and Sciences of Bilkent University.
On utilizing self-organizing fuzzy neural networks for financial forecasts in the NN5 forecasting competition Time series forecasting competition for computational intelligence The distance traveled to visit N points with a maximum of C stops per vehicle: An analytic model and an application
- D Coyle
- G Prasad
- T M Mcginnity
Coyle, D., Prasad, G., & McGinnity, T. M. (2010). On utilizing self-organizing fuzzy neural networks for financial forecasts in the NN5 forecasting competition. In Proceeding of: International joint conference on neural networks, IJCNN 2010, 18– 23 July. Barcelona, Spain. Crone, S. (2008). Time series forecasting competition for computational intelligence, <http://www.neural-forecastingcompetition.com>. Daganzo, C. F. (1984). The distance traveled to visit N points with a maximum of C stops per vehicle: An analytic model and an application. Transportation Science, 18(4), 331–350.
Cash demand forecasting for ATM using neural networks and support vector regression algorithms. In 20th EURO mini conference – continuous optimization and knowledge-based technologies
- R Simutis
- D Dilijonas
- L Vebastina
Simutis, R., Dilijonas, D., & veBastina, L. (2008). Cash demand forecasting for ATM using neural networks and support vector regression algorithms. In 20th EURO mini conference – continuous optimization and knowledge-based technologies, May 20–23 (pp. 416–421). Neringa, Lithuaniza.
Coke's RFID-based dispensers redefine business intelligence, information week
- M H Weier
Weier, M. H. (2009). Coke's RFID-based dispensers redefine business intelligence, information week, June 06, 2009. (http://www.informationweek.com//news/ mobility/RFID/217701971).