Article

# Traveling time prediction in scheduled transportation with journey segments

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

## Abstract

Urban mobility impacts urban life to a great extent. To enhance urban mobility, much research was invested in traveling time prediction: given an origin and destination, provide a passenger with an accurate estimation of how long a journey lasts. In this work, we investigate a novel combination of methods from Queueing Theory and Machine Learning in the prediction process. We propose a prediction engine that, given a scheduled bus journey (route) and a 'source/destination' pair, provides an estimate for the traveling time, while considering both historical data and real-time streams of information that are transmitted by buses. We propose a model that uses natural segmentation of the data according to bus stops and a set of predictors, some use learning while others are learning-free, to compute traveling time. Our empirical evaluation, using bus data that comes from the bus network in the city of Dublin, demonstrates that the snapshot principle, taken from Queueing Theory, works well yet suffers from outliers. To overcome the outliers problem, we use Machine Learning techniques as a regulator that assists in identifying outliers and propose prediction based on historical data.

## No full-text available

... There is very little discussion in the literature about the conceptual model of a bus route or network, although there is some evidence that it has an impact on the accuracy of predictions. The most common approach within the literature is to conceptualise a bus network as a series of routes made up of consecutive stop pair segments [12,17,38]. This natural segmentation of a bus route can have different implications depending on the number of stops on a bus route. ...
... The main trend in the literature to date is that while absolute error predictably increases while predicting for longer segments, the percentage (or relative error) decreases significantly [9,17,40,41]. Outside of bus journey time predictions, research on traffic prediction, in general, has also established that long term prediction is less susceptible to random disturbances and shows more regular patterns than short-term predictions [7]. We build on these findings with our approach and predict the whole journey time from origin to terminus stop in a given direction. ...
... Gal et al. [17] compare many different ML models, including RF on data from Dublin. RF performed well, especially with more data. ...
Full-text available
Article
Buses are a vital component of an urban environment and shifting away from private cars towards public transport is important in minimising our environmental impact and creating sustainable cities. Good bus services make urban life better and safer for everyone and having reliable journey time estimates is a crucial component of a good transit service. Many techniques have been developed to predict journey times, including historical averages, statistical approaches and more recently machine learning (ML) algorithms. Several research efforts have shown that predicting the travel time of a complete bus journey is more accurate than predicting partial journeys. We propose a method of predicting travel time for a whole bus journey using ML algorithms combined with a novel post prediction segmentation technique to provide an estimate of partial journey times. This novel approach proportions the journey dynamically based on historical averages for the relevant day of the week and time of day. The ML algorithms we used to predict a whole journey time are Random Forest (RF), Support Vector Machine (SVM) and k Nearest-Neighbor (kNN). Our approach is applied to one year of data from the city-wide bus network in Dublin. Our proportioning technique gives excellent results compared to a baseline of the ratio of stop pair segments on the partial journey compared to the whole journey. The best performing ML algorithm was RF which achieved 0.16 mean absolute percentage error (MAPE) and 158 seconds mean absolute error (MAE) with our approach compared to 0.42 MAPE and 245 seconds MAE with the baseline method. The results are especially relevant on shorter journeys and on routes with large data sets. Our method achieved 0.21 MAPE on short journeys of less than 10 stops compared to 0.78 with the baseline method. This is a significant result as short journeys are challenging to predict accurately. Of the ML algorithms used, kNN required the least resources to train, whereas SVM returned the prediction quickest and required the least space to store.
... In the past decades, by using historical data or online data (obtained by the global positioning system), various forecasting models and techniques have been proposed to predict bus travel time. These techniques include historical average model [4][5][6], statistical model [7][8][9][10][11], nonparametric regression model [2,[12][13][14], machine learning model [2,3,[15][16][17][18][19][20][21][22][23][24] and hybrid model [20,25,26]. ...
... In the past decades, by using historical data or online data (obtained by the global positioning system), various forecasting models and techniques have been proposed to predict bus travel time. These techniques include historical average model [4][5][6], statistical model [7][8][9][10][11], nonparametric regression model [2,[12][13][14], machine learning model [2,3,[15][16][17][18][19][20][21][22][23][24] and hybrid model [20,25,26]. ...
... Some applications of RFs can be found in bus travel time prediction. Gal et al. [20] proposed a combination method of queuing theory and RFs to predict bus travel time. In their paper, RFs were used to detect the outliers in historical data. ...
Full-text available
Article
In an intelligent transportation system, accurate bus information is vital for passengers to schedule their departure time and make reasonable route choice. In this paper, an improved deep belief network (DBN) is proposed to predict the bus travel time. By using Gaussian–Bernoulli restricted Boltzmann machines to construct a DBN, we update the classical DBN to model continuous data. In addition, a back-propagation (BP) neural network is further applied to improve the performance. Based on the real traffic data collected in Shenyang, China, several experiments are conducted to validate the technique. Comparison with typical forecasting methods such as k-nearest neighbor algorithm (k-NN), artificial neural network (ANN), support vector machine (SVM) and random forests (RFs) shows that the proposed method is applicable to the prediction of bus travel time and works better than traditional methods.
... Using historical or real-time data, multiple prediction methods and mechanisms have been developed to predict bus travel time. These methods can be categorized into statistical methods [8][9][10], machine learning methods [11][12][13][14], and neural network methods [1,5,7,15,16]. Statistical methods can be separated into historical average methods, regression methods and time-series methods [7]. ...
... They used data collected in Melbourne, Australia to validate the developed model. Gal et al. [13] combined a queuing theory with RFs, which were used to detect outliers in historical data, to predict bus travel time. Yu et al. [11] proposed an RFs model to predict bus travel time. ...
Full-text available
Article
Accurate travel time prediction allows passengers to schedule their journeys efficiently. However, cyclical factors (time intervals of the day, weather conditions, and holidays), unpredictable factors (incidents, abnormal weather), and other complicated factors (dynamic traffic conditions, dwell times, and variation in travel demand) make accurate bus travel time prediction complicated. This paper aims to achieve accurate travel time prediction. To do so, we propose a clustering method that identifies travel time paradigms of different route links and clusters them based on their similarity using the nonnegative matrix factorization algorithm. Additionally, we propose a deep learning model based on CNN with spatial–temporal attention and gating mechanisms to select the most relevant features and capture their dependencies and correlations. For each defined cluster, we train a separate model to predict the travel time at various time intervals over the day. As a result, the travel times of all journey links from related prediction models are aggregated to predict the total journey time. Extensive experiments using data collected from four different bus lines in Beijing show that our method outperforms the compared baselines.
... In 2017, Gal et al. [13] applied AdaBoost (AB) and GBM together with snapshot method from Queueing Theory to improve ensembles of regression trees for predicting travel time based on historical data of scheduled bus journals in Dublin, Ireland. Gal et al. [13] proved that combining snapshot rule with GBM leads to a more robust and better predictions regardless of the increase in trip length. ...
... In 2017, Gal et al. [13] applied AdaBoost (AB) and GBM together with snapshot method from Queueing Theory to improve ensembles of regression trees for predicting travel time based on historical data of scheduled bus journals in Dublin, Ireland. Gal et al. [13] proved that combining snapshot rule with GBM leads to a more robust and better predictions regardless of the increase in trip length. In the same year, Estes et al. [14] applied GBM and RF to solve delay prediction problem in air transport, and their work demonstrated that GBM outperformed RF. ...
Full-text available
Article
Passenger train delay significantly influences riders’ decision to choose rail transport as their mode choice. This article proposes real-time passenger train delay prediction (PTDP) models using the following machine learning techniques: random forest (RF), gradient boosting machine (GBM), and multi-layer perceptron (MLP). In this article, the impact on the PTPD models using Real-time based Data-frame Structure (RT-DFS) and Real-time with Historical based Data-frame Structure (RWH-DFS) is investigated. The results show that PTDP models using MLP with RWH-DFS outperformed all other models. The influence of the external variables such as historical delay profiles at the destination (HDPD), ridership, population, day of the week, geography, and weather information on the real-time PTPD models are also further analyzed and discussed.
... In the study [12], the prediction of the travel time was carried out using speed predictions. In [13], the prediction of the travel time for travel sections in planned transportation was studied. In [14], the prediction of the travel time was investigated with the help of GPS-based traveling vehicles. ...
... In [12], the PRD (percent root mean distortion) value of the prediction values is in the range of 0.1-0.001%. In [13], the R 2 value for the predictions produced was determined to be 0.35. In [14], it was stated that 42.86% success was achieved in the travel time predictions. ...
... However, both of them learn the linear weight of corresponding regression models, so they cannot handle the nonlinearity. To address this issue, other machine learning methods, such as DT (decision tree [130]) and HMM (hidden Markov model [131]), are applied to solve the problem. In particular, they partition the whole path into a sequence of links and then estimate each link's travel time. ...
... In particular, they partition the whole path into a sequence of links and then estimate each link's travel time. The authors in [130] independently estimate each link via some boosting techniques (such as AdaBoost and gradient boosting tree), while the authors in [131] model the whole sequence via the HMM technique. However, these traditional methods ignore some useful features, such as spatial and temporal properties. ...
Full-text available
Article
Intelligent transportation (e.g., intelligent traffic light) makes our travel more convenient and efficient. With the development of mobile Internet and position technologies, it is reasonable to collect spatio-temporal data and then leverage these data to achieve the goal of intelligent transportation, and here, traffic prediction plays an important role. In this paper, we provide a comprehensive survey on traffic prediction, which is from the spatio-temporal data layer to the intelligent transportation application layer. At first, we split the whole research scope into four parts from bottom to up, where the four parts are, respectively, spatio-temporal data, preprocessing, traffic prediction and traffic application. Later, we review existing work on the four parts. First, we summarize traffic data into five types according to their difference on spatial and temporal dimensions. Second, we focus on four significant data preprocessing techniques: map-matching, data cleaning, data storage and data compression. Third, we focus on three kinds of traffic prediction problems (i.e., classification, generation and estimation/forecasting). In particular, we summarize the challenges and discuss how existing methods address these challenges. Fourth, we list five typical traffic applications. Lastly, we provide emerging research challenges and opportunities. We believe that the survey can help the partitioners to understand existing traffic prediction problems and methods, which can further encourage them to solve their intelligent transportation applications.
... Our proposed WA predictor, which is a weighted sum of the LES announcement and EA, is in the same spirit as delay announcements which combine recent delay-history-based information along with average wait-time predictions, as in Gal et al. (2017) and Ang et al. (2015). However, our focus here is on analytically characterising the performance of the announcements, in addition Author: A correlation-based approach 6 Article submitted to Management Science; manuscript no. ...
... From a practical and empirical standpoint, there is some empirical evidence substantiating the good performance of the LES announcement with real-life data in some cases; e.g., see Senderovich et al. (2014), Senderovich et al. (2015), and Gal et al. (2017). In contrast, we begin here by presenting conflicting empirical evidence which illustrates the poor accuracy of the LES announcement in other cases, relative to the static announcement EA; we will return to this empirical evidence in §6.2.1. ...
Article
Service providers often share delay information, in the form of delay announcements, with their customers. In practice, simple delay announcements, such as average waiting times or a weighted average of previously delayed customers, are often used. Our goal in this paper is to gain insight into when such announcements perform well. Specifically, we compare the accuracies of two announcements: (i) a static announcement that does not exploit real-time information about the state of the system and (ii) a dynamic announcement, specifically the last-to-enter-service (LES) announcement, which equals the delay of the last customer to have entered service at the time of the announcement. We propose a novel correlation-based approach that is theoretically appealing because it allows for a comparison of the accuracies of announcements across different queueing models, including multiclass models with a priority service discipline. It is also practically useful because estimating correlations is much easier than fitting an entire queueing model. Using a combination of queueing-theoretic analysis, real-life data analysis, and simulation, we analyze the performance of static and dynamic announcements and derive an appropriate weighted average of the two which we demonstrate has a superior performance using both simulation and data from a call center. This paper was accepted by Vishal Gaur, operations management.
... Most solutions on travel time estimation require information on the traveled route [5,6,7,8,9]. Although applicable in the real world, they introduce uncertainty when requiring a pre-planned route for prediction, which in practice may be impossible. ...
... Given hour-ofday, day-of-week, and weather conditions (precipitation), the network would estimate the travel time for each route segment, further adjusting the prediction using Kalman filtering and up-to-the-minute bus locations. Similarly, the work in [6] estimates the duration of a bus ride based on a scheduled route and a source-destination pair. In [7], travel time was predicted based on floating-car data, by combining linear models, deep, and recurrent neural networks to explore road segment sequences. ...
Preprint
The acquisition of massive data on parcel delivery motivates postal operators to foster the development of predictive systems to improve customer service. Predicting delivery times successive to being shipped out of the final depot, referred to as last-mile prediction, deals with complicating factors such as traffic, drivers' behaviors, and weather. This work studies the use of deep learning for solving a real-world case of last-mile parcel delivery time prediction. We present our solution under the IoT paradigm and discuss its feasibility on a cloud-based architecture as a smart city application. We focus on a large-scale parcel dataset provided by Canada Post, covering the Greater Toronto Area (GTA). We utilize an origin-destination (OD) formulation, in which routes are not available, but only the start and end delivery points. We investigate three categories of convolutional-based neural networks and assess their performances on the task. We further demonstrate how our modeling outperforms several baselines, from classical machine learning models to referenced OD solutions. Specifically, we show that a ResNet architecture with 8 residual blocks displays the best trade-off between performance and complexity. We perform a thorough error analysis across the data and visualize the deep features learned to better understand the model behavior, making interesting remarks on data predictability. Our work provides an end-to-end neural pipeline that leverages parcel OD data as well as weather to accurately predict delivery durations. We believe that our system has the potential not only to improve user experience by better modeling their anticipation but also to aid last-mile postal logistics as a whole.
... Every bus transmits its timestamped location at a 20-seconds resolution, which also includes information on the nearest bus stop. The readings are processed in real-time, thus allowing for online analysis, such as anomaly detection and travelling time prediction [26], [17]. ...
... Our model enables us to effectively use the recorded data to answer complex queries, without the need to assume 'continuous tracking' (no missing events and timestamps) [26], or a bias by imputation of missing timestamps [17]. Analysis of the transportation network in Dublin is based on a stop congestion query, testing the co-location of buses at specific stops. ...
Article
Recognising patterns that correlate multiple events over time becomes increasingly important in applications that exploit the Internet of Things, reaching from urban transportation through surveillance monitoring to business workflows. In many real-world scenarios, however, timestamps of events may be erroneously recorded, and events may be dropped from a stream due to network failures or load shedding policies. In this work, we present SimpMatch, a novel simplex-based algorithm for probabilistic evaluation of event queries using constraints over event orderings in a stream. Our approach avoids learning probability distributions for time-points or occurrence intervals. Instead, we employ the abstraction of segmented intervals and compute the probability of a sequence of such segments using the notion of order statistics. The algorithm runs in linear time to the number of lost events and shows high accuracy, yielding exact results if event generation is based on a Poisson process and providing a good approximation otherwise. We demonstrate empirically that SimpMatch enables efficient and effective reasoning over event streams, outperforming state-of-the-art methods for probabilistic evaluation of event queries by up to two orders of magnitude.
... Most of the recent predictions of travel time have adopted machine learning methods [13,14] including k-nearest neighbor (KNN) algorithms [15,16], support vector machines [17], and neural networks [18,19] model. Compared to earlier statistical prediction methods, machine learning models do not assume any specific model structure for the data, but treat it as unknown, which can handle complex problems and large amounts of data well. ...
... In equation (14), y 0 , y, y 1 is a series of data points arranged in ascending order from time to time, which is the previous point closest to the missing point y in time and x 0 is the corresponding time point. Similarly, y 1 is the next point closest to y in time and x 1 is the corresponding time point. ...
Full-text available
Article
With the rapid growth of car ownership, traffic congestion has become one of the most serious social problems. For us, accurate real-time travel time predictions are especially important for easing traffic congestion, enabling traffic control and management, and traffic guidance. In this paper, we propose a method to predict urban road travel time by combining XGBoost and LightGBM machine learning models. In order to obtain a relatively complete data set, we mine the GPS data of Beijing and combine them with the weather feature to consider the obtained 14 features as candidate features. By processing and analyzing the data set, we discussed in detail the correlation between each feature and the travel time and the importance of each feature in the model prediction results. Finally, the 10 important features screened by the LightGBM and XGBoost models were used as key features. We use the full feature set and the key feature set as input to the model to explore the effect of different feature combinations on the prediction accuracy of the model and then compare the prediction results of the proposed fusion model with a single model. The results show that the proposed fusion model has great advantages to urban travel time prediction.
... In the proposed methods, SVM is combined with a Genetic Algorithm [5], Kalman filter [6], and artificial neural network (ANN) [7], respectively. In addition to Kalman filtering and SVM, there are other time series prediction methods, such as road segment average travel time [8], the Relevance Vector Machine Regression [9], clustering [10], Queueing Theory combined with Machine Learning [11], and Random Forests [12]. Artificial neural networks have been widely used in various research fields in recent years [13][14][15]. ...
... 11A long Short-Term Memory (LSTM) unit structure. ...
Full-text available
Article
Accurate prediction of bus arrival times is a challenging problem in the public transportation field. Previous studies have shown that to improve prediction accuracy, more heterogeneous measurements provide better results. So what other factors should be added into the prediction model? Traditional prediction methods mainly use the arrival time and the distance between stations, but do not make full use of dynamic factors such as passenger number, dwell time, bus driving efficiency, etc. We propose a novel approach that takes full advantage of dynamic factors. Our approach is based on a Recurrent Neural Network (RNN). The experimental results indicate that a variety of prediction algorithms (such as Support Vector Machine, Kalman filter, Multilayer Perceptron, and RNN) have significantly improved performance after using dynamic factors. Further, we introduce RNN with an attention mechanism to adaptively select the most relevant input factors. Experiments demonstrate that the prediction accuracy of RNN with an attention mechanism is better than RNN with no attention mechanism when there are heterogeneous input factors. The experimental results show the superior performances of our approach on the data set provided by Jinan Public Transportation Corporation.
... Attempts in the literature exist to predict patient traffic, but most of the work focuses on patient flow in the emergency department (ED) [15,16]. Furthermore, independent attempts have been made to predict and schedule traffic flow [17,18]. However, to our knowledge, no literature has explored the integration of the two processes using machine learning techniques. ...
... From Equation (18) above, the nearest bus stop will be the one with the shortest distance. The live location of a bus is queried based either on target destination or bus number. ...
Full-text available
Article
A mismatch between staffing ratios and service demand leads to overcrowding of patients in waiting rooms of health centers. Overcrowding consequently leads to excessive patient waiting times, incomplete preventive service delivery and disgruntled medical staff. Worse, due to the limited patient load that a health center can handle, patients may leave the clinic before the medical examination is complete. It is true that as one health center may be struggling with an excessive patient load, another facility in the vicinity may have a low patient turn out. A centralized hospital management system, where hospitals are able to timely exchange patient load information would allow excess patient load from an overcrowded health center to be re-assigned in a timely way to the nearest health centers. In this paper, a machine learning-based patient load prediction model for forecasting future patient loads is proposed. Given current and historical patient load data as inputs, the model outputs future predicted patient loads. Furthermore, we propose re-assigning excess patient loads to nearby facilities that have minimal load as a way to control overcrowding and reduce the number of patients that leave health facilities without receiving medical care as a result of overcrowding. The re-assigning of patients will imply a need for transportation for the patient to move from one facility to another. To avoid putting a further strain on the already fragmented ambulatory services, we assume the existence of a scheduled bus system and propose an Internet of Things (IoT) integrated smart bus system. The developed IoT system can be tagged on buses and can be queried by patients through representation state transfer application program interfaces (APIs) to provide them with the position of the buses through web app or SMS relative to their origin and destination stop. The back end of the proposed system is based on message queue telemetry transport, which is lightweight, data efficient and scalable, unlike the traditionally used hypertext transfer protocol.
... The large amount of information obtained on a daily basis requires a specialized evaluation to result in improvements in the service. Following this section, several papers [6,7] were published focusing on the analysis of the dynamic data obtained to learn how the population uses public transport and thereby improve the service provided. These studies assume that passenger counting is effective, but as reported by Oberli [8] there are a number of factors that generate losses and therefore these numbers do not describe the actual scenario of use. ...
... Seeking a better performance in reading, this article is focusing in two specific aspects: the analysis of tags options On the left hand (2) In the front left pocket of pants (3) In the front right pocket of pants (4) In the right sock (5) In the shirt pocket (6) In the left sock (7) On the right hand (8) In the right front pocket of pants along with keys (9) In the right front pocket of pants along with the cell phone (10) In the back right pocket of pants (11) In the back left pocket of pants (12) Inside the wallet in the back right pocket of pants (13) In the backpack (14) In the wallet inside the backpack The results obtained from experimental tests show how and where the readers' antennas should be allocated, as well as which tag options are the less susceptible to interference from the human body. With this information it is possible to imagine a better application of RFID in the user detection as suggested by the AFC. ...
Full-text available
Article
Automatic control for recognition of passengers in public transport systems has been a crucial point in mobility systems towards enhancements in passengers’ flow and overall system efficiency. It allows the recognition of passengers’ origins and destinations, so that the specific demands for specific periods of the day can be assessed for an effective system planning. However, this automatic control has to be efficient and smooth so that it does not incur in additional overhead to the entire system. This work presents a study on a passenger recognition system for public transport through the use of RFID technology using EPC Gen2 standard. Preliminary tests were performed with two different forms of voluntary order to evaluate different types of tags. These tests first evaluated the height and angle of the antennas using 1, 2, 3 and 4 antennas in the tag recognition. From the results of these first tests, a set up was defined and then applied to a second evaluation now with 10 volunteers, which evaluated repeatability and effectiveness of the system for recognition. Moreover, additional laboratory-based tests were performed to access the effectiveness of the proposed recognition system. The acquired results provide a basis for evaluate the suitability and applicability of the proposed system.
... Bus travel time on a particular path has time sequence characteristics (i.e. consecutive buses operate in a similar traffic condition) [25]. However, the SVM model has limitations to forecast time sequence information, such as bus travel time. ...
Full-text available
Article
The application of predicting bus travel time with real-time information, including Global Positioning System (GPS) and Electronic Smart Card (ESC) data is effective to advance the level of service by reducing wait time and improving schedule adherence. However, missing information in the data stream is inevitable for various reasons, which may seriously affect prediction accuracy. To address this problem, this research proposes a Long Short-Term Memory (LSTM) model to predict bus travel time, considering incomplete data. To improve the model performance in terms of accuracy and efficiency, a Genetic Algorithm (GA) is developed and applied to optimise hyperparameters of the LSTM model. The model performance is assessed by simulation and real-world data. The results suggest that the proposed approach with hybrid data outperforms the approaches with ESC and GPS data individually. With GA, the proposed model outperforms the traditional one in terms of lower Root Mean Square Error (RMSE). The prediction accuracy with various combinations of ESC and GPS data is assessed. The results can serve as a guideline for transit agencies to deploy GPS devices in a bus fleet considering the market penetration of ESC.
... As future work, we plan to include traffic flow data measured by sensor stations and to incorporate queueing theory and machine learning techniques as in Gal et al. (2017) for traveling time prediction. Other health indicators or emergency situations, such as traffic accidents, could be included in our FIS, or in the final decision made by the domain expert to close an area. ...
Full-text available
Article
The control of the pandemic caused by SARS-CoV-2 is a challenge for governments all around the globe. To manage this situation, countries have adopted a bundle of measures, including restrictions to population mobility. As a consequence, drivers face with the problem of obtaining fast routes to reach their destinations. In this context, some recent works combine Intelligent Transportation Systems (ITS) with big data processing technologies taking the traffic information into account. However, there are no proposals able to gather the COVID-19 health information, assist in the decision-making process, and compute fast routes in an all-in-one solution. In this paper, we propose a Pandemic Intelligent Transportation System (PITS) based on Complex Event Processing (CEP), Fuzzy Logic (FL) and Colored Petri Nets (CPN). CEP is used to process the COVID-19 health indicators and FL to provide recommendations about city areas that should not be crossed. CPNs are then used to create map models of health areas with the mobility restriction information and obtain fast routes for drivers to reach their destinations. The application of PITS to Madrid region (Spain) demonstrates that this system provides support for authorities in the decision-making process about mobility restrictions and obtain fast routes for drivers. PITS is a versatile proposal which can easily be adapted to other scenarios in order to tackle different emergency situations.
... Additionally, recent studies have shown that it is possible to precise and accurately estimate the duration of various processes in multiple fields. In this regard, several studies are noteworthy, including [3][4][5][6][7][8][9][10][11][12][13], which used machine learning techniques to estimate the duration of different processes in different fields. Such studies typically estimate the time spent using extensive multidimensional time series datasets [12], which are, in short, a collection of correlated observations made sequentially over time. ...
Full-text available
Article
Brazilian legal system prescribes means of ensuring the prompt processing of court cases, such as the principle of reasonable process duration, the principle of celerity, procedural economy, and due legal process, with a view to optimizing procedural progress. In this context, one of the great challenges of the Brazilian judiciary is to predict the duration of legal cases based on information such as the judge, lawyers, parties involved, subject, monetary values of the case, starting date of the case, etc. Recently, there has been great interest in estimating the duration of various types of events using artificial intelligence algorithms to predict future behaviors based on time series. Thus, this study presents a proof-of-concept for creating and demonstrating a mechanism for predicting the amount of time, after the case is argued in court (time when a case is made available for the magistrate to make the decision), for the magistrate to issue a ruling. Cases from a Regional Labor Court were used as the database, with preparation data in two ways (original and discretization), to test seven machine learning techniques (i) Multilayer Perceptron (MLP); (ii) Gradient Boosting; (iii) Adaboost; (iv) Regressive Stacking; (v) Stacking Regressor with MLP; (vi) Regressive Stacking with Gradient Boosting; and (vii) Support Vector Regression (SVR), and determine which gives the best results. After executing the runs, it was identified that the adaboost technique excelled in the task of estimating the duration for issuing a ruling, as it had the best performance among the tested techniques. Thus, this study shows that it is possible to use machine learning techniques to perform this type of prediction, for the test data set, with an R ² of 0.819 and when transformed into levels, an accuracy of 84%.
... Unsupervised clustering algorithms can be selected to distinguish these transfer choices, which include linear discriminant analysis (Krygsman, Dijst, and Arentze 2004), artificial neural network (Garrido, De Oña, and De Oña 2014), support vector machine (Yu et al. 2017), and decision tree (Hernandez, Monzon, and de Oña 2016). Typical applications of unsupervised algorithms on similar problems include the usage of principal components analysis (Lois, Monzón, and Hernández 2018), random forests (Gal et al. 2017), and K-means clustering (Zhang et al. 2019). Semi-supervised methods may be a better choice when a small amount of labeled data is available. ...
Full-text available
Article
This study presents a comprehensive framework for estimating passengers' transfer times and extracting their distribution and related transfer routes using WIFI probe data. The departure time of preceding station, arrival time of subsequent station, and train running time are selected to obtain transfer times. Then, the collected data is analyzed using kernel density estimation to obtain candidate distribution. Gaussian mixture models are adopted to extract the distribution of each possible transfer route at both peak hours and off-peak hours. This method is tested at two transfer stations of Xi’an metro system with the comparison of results from automatic fare collection data and manual sampling survey data. The results indicate that the proposed approach can collect the transfer time with a sampling ratio greater than 30% and a deviation less than 5%. The route choice behaviors and distribution of transfer time under various conditions can be identified using the proposed methods.
... In this dataset, the travel times are 15-minute interval average travel times for each route. Gal et al. [4] investigate methods from Queueing Theory and Machine Learning in the prediction process. They implemented model based on segmentation of the travel time into stop-based segments using bus data of Dublin city. ...
... There are numerous machine (and deep) learning solutions for the food industry domain [28,37,63] and other domains with similar characteristics, such as transportation [18,51] and air pollution [23,52]. Common to all of these works is the availability of labeled data. ...
Full-text available
Preprint
Industry 4.0 offers opportunities to combine multiple sensor data sources using IoT technologies for better utilization of raw material in production lines. A common belief that data is readily available (the big data phenomenon), is oftentimes challenged by the need to effectively acquire quality data under severe constraints. In this paper we propose a design methodology, using active learning to enhance learning capabilities, for building a model of production outcome using a constrained amount of raw material training data. The proposed methodology extends existing active learning methods to effectively solve regression-based learning problems and may serve settings where data acquisition requires excessive resources in the physical world. We further suggest a set of qualitative measures to analyze learners performance. The proposed methodology is demonstrated using an actual application in the milk industry, where milk is gathered from multiple small milk farms and brought to a dairy production plant to be processed into cottage cheese.
... Furthermore, after considering several machine learning methods, they selected artificial neural networks for the job. In a similar way, Gal et al. [40] considered both historical and real-time data associated with the bus network system in the city of Dublin to develop a hybrid method combining queuing theory and machine learning. They employ this method to predict travel times in scheduled bus routes. ...
Full-text available
Article
With the emergence of fog and edge computing, new possibilities arise regarding the data-driven management of citizens’ mobility in smart cities. Internet of Things (IoT) analytics refers to the use of these technologies, data, and analytical models to describe the current status of the city traffic, to predict its evolution over the coming hours, and to make decisions that increase the efficiency of the transportation system. It involves many challenges such as how to deal and manage real and huge amounts of data, and improving security, privacy, scalability, reliability, and quality of services in the cloud and vehicular network. In this paper, we review the state of the art of IoT in intelligent transportation systems (ITS), identify challenges posed by cloud, fog, and edge computing in ITS, and develop a methodology based on agile optimization algorithms for solving a dynamic ride-sharing problem (DRSP) in the context of edge/fog computing.These algorithms allow us to process, in real time, the data gathered from IoT systems in order to optimize automatic decisions in the city transportation system, including: optimizing the vehicle routing, recommending customized transportation modes to the citizens, generating efficient ride-sharing and car-sharing strategies, create optimal charging station for electric vehicles and different services within urban and interurban areas. A numerical example considering a DRSP is provided, in which the potential of employing edge/fog computing, open data, and agile algorithms is illustrated.
... e travel time of bus contains running time and stopping time. Studies show that the running time is affected by traffic conditions of roads and traffic signals [46,47]. Kieu et al. analyzed the distribution of transit travel time used transit signal priority data and found that they followed lognormal distributions [48]. ...
Full-text available
Article
Operational efficiency and stability are two critical aspects to measure bus systems. Influenced by many stochastic factors, buses always suffer from delay and bunching. Traditional studies focus on a single route and lack research on the systematic evaluation of bus network. In this paper, we propose a data-driven framework to analyze the efficiency and stability based on small granularity GPS trajectory data from the perspective of entire bus network. The IC card data and route data are used to extract the boarding passenger number and topological structure, respectively. The results show that the average headway of stations follows a lognormal distribution. Moreover, the distribution of arrival efficiency of stations is inhomogeneous and a small number of stations have large values. In addition, the relationships among average headway of stations, boarding passenger number, bus number, and complex network indicators are revealed. It is found that the average headway of station is negatively correlated with other indicators, which implies that complex network connections and more passenger flows could weaken the efficiency of bus operations. This paper provides a way to evaluate the operational performance of bus networks and could give help for monitoring and optimizing the daily operation of bus systems.
... The authors used a measure of activity density to measure and map household and traffic congestion trends in space. Avigdor Gal et al. [28] have proposed a model that combines both Queueing Theory and Machine Learning techniques. The authors define the natural segmentation of the data according to intermediate stops. ...
... The authors used a measure of activity density to measure and map household and traffic congestion trends in space. Avigdor Gal et al. [28] have proposed a model that combines both Queueing Theory and Machine Learning techniques. The authors define the natural segmentation of the data according to intermediate stops. ...
Full-text available
Article
With the rapid expansion of sensor technologies and wireless network infrastructure, research and development of traffic associated applications, such as real-time traffic maps, on-demand travel route reference and traffic forecasting are gaining much more attention than ever before. In this paper, we elaborate on our traffic prediction application, which is based on traffic data collected through Google Map API. Our application is a desktop-based application that predicts traffic congestion state using Estimated Time of Arrival (ETA). In addition to ETA, the prediction system takes into account various features such as weather, time period, special conditions, holidays, etc. The label of the classifier is identified as one of the five traffic states i.e. smooth, slightly congested, congested, highly congested or blockage. The results demonstrate that the random forest classification algorithm has the highest prediction accuracy of 92 percent followed by XGBoost and KNN respectively.
... is may lead to the redesign of routes or part of the network to improve the transport system. ere has been a considerable amount of research into designing effective approaches so that accurate and adaptive prediction systems can be offered to both bus companies and passengers [37][38][39]. Historical data, such as arrival times against planned times for given stops over an extended period, are essential to provide a deep insight into the behaviour of a transport network. Combined with machine learning techniques, such data allow forecasting about the network's status, even when there is no information available about external factors [23,[40][41][42][43]. Existing systems have exploited neural networks and regression and clustering techniques to predict bus arrival times [43][44][45][46][47], while others have performed route prediction using GPS data observation [48]. ...
Full-text available
Article
Efficient management of smart transport systems requires the integration of various sensing technologies, as well as fast processing of a high volume of heterogeneous data, in order to perform smart analytics of urban networks in real time. However, dynamic response that relies on intelligent demand-side transport management is particularly challenging due to the increasing flow of transmitted sensor data. In this work, a novel smart service-driven, adaptable middleware architecture is proposed to acquire, store, manipulate, and integrate information from heterogeneous data sources in order to deliver smart analytics aimed at supporting strategic decision-making. The architecture offers adaptive and scalable data integration services for acquiring and processing dynamic data, delivering fast response time, and offering data mining and machine learning models for real-time prediction, combined with advanced visualisation techniques. The proposed solution has been implemented and validated, demonstrating its ability to provide real-time performance on the existing, operational, and large-scale bus network of a European capital city.
... A few years later, Senderovich et al. experimented on another medical dataset, with RTLS data taken from an American hospital, with approximately 240,000 events per year [5]. In 2014, a smart city dataset, which was used in a task that analyzed bus routes as processes contained over 1 million events per day for a period of one month (approximately 30 million events) [6]. Data velocity has been a main focus of process mining, not due to the speed of event arrival (after all, registering a new child in a Dutch municipality does not have to be done in a matter of seconds) but due to the inherent notion of change, which is common to data velocity and the understanding of the nature of a process. ...
Chapter
The discipline of process mining was inaugurated in the BPM community. It flourished in a world of small(er) data, with roots in the communities of software engineering and databases and applications mainly in organizational and management settings. The introduction of big data, with its volume, velocity, variety, and veracity, and the big strides in data science research and practice pose new challenges to this research field. The paper positions process mining along modern data life cycle, highlighting the challenges and suggesting directions in which data science disciplines (e.g., machine learning) may interact with a renewed process mining agenda.
... The HTTP framework described by Wang-Chien et al. [11] uses historical data by identifying and grouping similar trajectories and then combines two hybrid prediction schemes which outperforms state-of-the-art schemes. Whereas Gal et al. [12] segments a bus journey and uses a combination of Machine Learning and Queueing Theory predictors to model travelling time in each segment. They find that the length of the journey does not influence the predictions and supports applying mixed Queue and Machine Learning predictors in similar settings. ...
Full-text available
Experiment Findings
Travel time is a vital component of road-based transit networks and delays are a major aspect that affect this transit time. This paper focuses on building a system that collects transit feed in real time and uses that information to predict delays. The predicted delays, along with attributes from transit feed, are used to group delays into classes based on similarities and then further maps these classes to causes. We also try to compare the quality of various forecast and classification models. This study is based on real-time bus transit information published by Dublin Bus. The experimental data used in this analysis was collected using java/spring based batch jobs scheduled based on publicly available bus transit schedule. The collected data, which amounts to about a quarter million records a day, are then pooled in to Google BigQuery, where the data is cleaned and preprocessed. The data collected roughly estimates to be about 5 million records and growing. The delays are then predicted using models based on regression, ensemble and ANNs, and the results are compared. The similarity of patterns in the predicted delays, along with other attributes, are then used to group these delay instances to various probable groups using unsupervised clustering methods like K Means, DBScan and OPTICS algorithms, and these groups are then labelled with causes. The cause-groups are then used to train classification models to predict the cause of new incoming real time transit delay record.
... The use of a deep architecture model was enhanced by the representation of traffic flow features through autoencoders byLv et al. (2015). A hybrid framework using Queueing Theory and ML was used to address outliers in an application of travel time prediction byGal et al. (2017).It is worth noting that the aforementioned studies did not account for the spatial dimension, which has started receiving attention very recently due to advances in computing power.Yu et al. (2017) used a spatiotemporal recurrent convolutional network (SRCN), which inherits the advantages of deep CNNs and long short-term memory (LSTM) NNs for traffic forecasting. Incorporating the spatial domain resulted in better performance compared to traditional deep architectures for both short-and long-term forecasts. ...
Thesis
Emerging transportation technologies like autonomous vehicles and services like on-demand shared mobility are casting their shadows over the traditional paradigm of vehicle ownership. Several countries are witnessing stagnation in overall car use, perhaps due to the proliferation of access-based services and changing attitudes of millennials. Therefore, it becomes necessary to revisit this paradigm, and reconsider strategies for modeling vehicle availability and use in this new era. This thesis attempts to do that through three studies that contribute to the methodological, conceptual, and praxis literatures. The first study proposes a hybrid modeling methodology that leverages machine learning techniques to enhance traditional behavioral discrete choice models used in practice. The usefulness of this model to predict market shares of unforeseen choices like new mobility services is illustrated through an application to the off-peak car in Singapore. Our model significantly improves upon the market shares predicted by traditional models through an average reduction of 60% in RMSE. The second study shifts the focus from vehicle ownership to vehicle availability in the form of mobility bundles. We leverage Singapore’s unique policy environment to empirically model households’ preferences for unique mobility bundles that are constructed in an ordinal fashion. This is followed by an examination of car usage within the household. Significant intra-household interaction effects are found with respect to job location, in addition to the observation of gender biases in the decision-making process. The third study evaluates the effectiveness of car-lite policies that seek to replace private vehicle usage with shared and smart mobility services. Behavioral responses to the policy and associated market effects are modeled using an integrated land use transport simulator calibrated for Singapore. Initially favorable aggregate outcomes tend to disappear as short-term market effects set in. Although outcomes stabilize to a certain extent over the long-term, the initial characteristics of the study area are found to strongly influence the success of such policies.
... In order to effectively cope with the influences of the incidents in the bus arrival prediction problem, this paper proposes a forecasting analysis method based on space feature vector. In this method, the bus driving path is divided into several segments according to the intersection of the road and calculates the current speed of all buses on each road [2]. Finally, the weighted average is calculated as the instantaneous speed of the bus at the current section, and the required time for the bus to pass the current section is calculated which will be treated as the eigenvalue of the spatial vector during the stage of prediction. ...
Full-text available
Article
Bus arrival prediction has important implications for public travel, urban dispatch, and mitigation of traffic congestion. The factors affecting urban traffic conditions are complex and changeable. As the predicted distance increases, the difficulty of traffic prediction becomes more difficult. Forecast based on historical data responds quite slowly for changes under the short-term conditions, and vehicle prediction based on real-time speed is not sufficient to predict under long-term conditions. Therefore, an arrival prediction method based on temporal vector and another arrival prediction method based on spatial vector is proposed to solve the problems of remote dependence of bus arrival and road incidents, respectively. In this paper, combining the advantages of the two prediction models, this paper proposes a long short-term memory (LSTM) and Artificial neural networks (ANN) comprehensive prediction model based on spatial-temporal features vectors. The long-distance arrival-to-station prediction is realized from the dimension of time feature, and the short-distance arrival-to-station prediction is realized from the dimension of spatial feature, thereby realizing the bus-to-station prediction. Besides, experiments were conducted and tested based on the entity dataset, and the result shows that the proposed method has high accuracy among bus arrival prediction problems.
... Some researchers recognizes that several routes can benefit from each other's predictions if they share some partial route segments (Bai et al., 2015;Gal et al., 2017;Yu et al., 2011), but in all cases it is a prerequisite that the routes indeed has identical links (and thus e.g. identical stop point sequences) in all routes. ...
Full-text available
Article
Accurate and reliable predictions for bus arrival in public transport networks are essential for delivering an attractive service. This paper presents a multi-model approach for bus arrival prediction. The approach uses three distinct sub-models in an ensemble model. A multi-output, multi-time-step, deep neural network using Convolutional and Long short-term memory (LSTM) layers is used for travel time, and more simplistic models are used for dwell time and seasonal components. The method is empirically evaluated and compared to other popular approaches. We find that the proposed model saturations outperforms the other methods, while in other saturations performs similar.
... Moreira-Matias et al. [2] conduct a comprehensive review of techniques used for this type of prediction. Focusing on Machine Learning techniques, taking into account the type of technique used and the references number, we highlight Yu et al. [3] who proposed models based on support vector machines, Bai et al. [4] who proposed a combined model based on support vector machine and Kalman filters, Gurmu et al. [5] who presented a prediction model based on artificial neuronal networks, Chang et al. [6] who proposed the technique k-nearest neighbors, Gal et al. [7] that used decision tree regression and finally, the work of Lee et al. [8] that proposed clustering techniques, specifically K-means and V-means. All these short-term TT forecasting models use, as input data, a set of TT observed at different points in the transport network in certain instants in time. ...
Full-text available
Article
In road-based mass transit systems, the travel time is a key factor affecting quality of service. For this reason, to know the behavior of this time is a relevant challenge. Clustering methods are interesting tools for knowledge modeling because these are unsupervised techniques, allowing hidden behavior patterns in large data sets to be found. In this contribution, a study on the utility of different clustering techniques to obtain behavior pattern of travel time is presented. The study analyzed three clustering techniques: K-medoid, Diana, and Hclust, studying how two key factors of these techniques (distance metric and clusters number) affect the results obtained. The study was conducted using transport activity data provided by a public transport operator.
... The correlations between the travel times of nearby links and different time slots are crucial for inferring the traffic state of a particular link [Niu et al. 2014;Zhang et al. 2016]. Online methods that determine the time required by a bus to reach a specified bus stop were proposed in [Gal et al. 2017], [Gal et al. 2018] and [Yu et al. 2011]. In [Wang et al. 2016b] the authors propose a method that estimates the travel time by identifying near-neighbor trajectories, with similar origin and destination. ...
Conference Paper
Travel time estimation is a critical task, useful to many urban applications at the individual citizen and the stakeholder level. This paper presents a novel hybrid algorithm for travel time estimation that leverages historical and sparse real-time trajectory data. Given a path and a departure time we estimate the travel time taking into account the historical information, the real-time trajectory data and the correlations among different road segments. We detect similar road segments using historical trajectories, and use a latent representation to model the similarities. Our experimental evaluation demonstrates the effectiveness of our approach.
... Previous work has shown that congestion has a substantial impact on the total time spent in a system (Gal et al. 2017) and hence on the quality of time prediction. However, event logs lack explicit information on the load imposed by arriving entities that are processed by shared (and scarce) resources. ...
Full-text available
Article
Time prediction is an essential component of decision making in various Artificial Intelligence application areas, including transportation systems, healthcare, and manufacturing. Predictions are required for efficient resource allocation and scheduling, optimized routing, and temporal action planning. In this work, we focus on time prediction in congested systems, where entities share scarce resources. To achieve accurate and explainable time prediction in this setting, features describing system congestion (e.g., workload and resource availability), must be considered. These features are typically gathered using process knowledge, (i.e., insights on the interplay of a system’s entities). Such knowledge is expensive to gather and may be completely unavailable. In order to automatically extract such features from data without prior process knowledge, we propose the model of congestion graphs, which are grounded in queueing theory. We show how congestion graphs are mined from raw event data using queueing theory based assumptions on the information contained in these logs. We evaluate our approach on two real-world datasets from healthcare systems where scarce resources prevail: an emergency department and an outpatient cancer clinic. Our experimental results show that using automatic generation of congestion features, we get an up to 23% improvement in terms of relative error in time prediction, compared to common baseline methods. We also detail how congestion graphs can be used to explain delays in the system.
... The structure of the P-LSTM model has two stages, as shown in Fig. 2. The first stage is the speculation of bus arrival times and passenger flows at stations, which includes inferring the number of boarding, alighting, and transferring passengers along with the bus arrival interval. The second stage is the prediction of passenger flow [39]- [43] and arrival times [44]- [47], which predicts the arrival time, boarding, alighting and transfer passengers using a Multilayer LSTM. ...
Full-text available
Article
Public transport is vital to people’s daily travel, and bus dispatching plays a significant role in the public transport system. With deep learning having been widely applied and achieved great success in many fields, bus dispatching methods based on deep learning are proposed in succession. Currently, many bus dispatching models assume that the bus departure timetable is fixed and optimize the bus departure timetable interval according to passenger flow. However, the bus departure timetable is variable in general, only considering that the bus arrival time is insufficient. Targeting the above challenges, we propose a novel dynamic bus dispatching model based on arrival time and passenger flow prediction (D-ATPF). First, the historical origin–destination (OD) data and the transfer data are obtained by processing the bus trajectory data and the passenger card-swiping records, and the bus arrival time is extracted by analyzing the GPS trajectory. Second, the components of bus arrival time and passenger flow prediction based on long short-term memory (P-LSTM) are adopted to predict the future passenger flow and bus arrival time. Finally, the genetic algorithm-based bus dispatching model (GABD model) searches the minimum waiting time for passengers by using stay strategy. By using data of five lines with 124 bus stations and a total of 9 02 509 records in Guangzhou city, China, our experimental results show that: 1) the average mean absolute percentage error ( $\overline {MAPE}$ ) and root mean square error ( $\overline {RMSE})$ of passenger prediction are 14% and 7.5, respectively; 2) the average $\overline {MAPE}$ and $\overline {RMSE}$ of bus arrival time are 7.5% and 13.5, respectively; 3) regarding the passenger flow and arrival time prediction, the proposed D-ATPF model reduced waiting time by 829.68 min, accounting for 25.19% of the total waiting time; and 4) compared with the real-time stay strategy, the reduced waiting time of this method increased by 5.94%. Therefore, the D-ATPF model provided a more practical model for buses dispatching.
... Tree ensemble method Evaluation scenario (Hamner, 2010) R F N o (Leshem and Ritov, 2007) adaptive boosting trees (AB) þ RF Effect of different replacement strategy for missing values (Mendes-Moreira et al., 2012) RF Effect of different feature input domain value definition (Zhang and Haghani, 2015) RF, GBRT Effect of different prediction horizon Effect of different traffic condition (Gal et al., 2017) RF, extremely randomized trees, AB, GBRT, GBRT-LAD a , Snapshot b þAB, SnapshotþGBRT, SnapshotþGBRT-LAD Effect of different trip length Effect of time-of-day, effect of bus load Missing data were taken into account when generating data, but no related discussion are found (Li and Bai, 2016a) GBRT Effect of different real-time feature inputs ...
Article
Purpose Many transport and logistics companies nowadays use raw vehicle GPS data for travel time prediction. However, they face difficult challenges in terms of the costs of information storage, as well as the quality of the prediction. This paper aims to systematically investigate various meta-data (features) that require significantly less storage space but provide sufficient information for high-quality travel time predictions. Design/methodology/approach The paper systematically studied the combinatorial effects of features and different model fitting strategies with two popular decision tree ensemble methods for travel time prediction, namely, random forests and gradient boosting regression trees. First, the investigation was conducted using pseudo travel time data that were generated using a pseudo travel time sampling algorithm, which allows generating travel time data using different noise processes so that the prediction performance under different travel conditions and noise characteristics can be studied systematically. The results and findings were then further compared and evaluated through a real-life case. Findings The paper provides empirical insights and guidelines about how raw GPS data can be reduced into a small-sized feature vector for the purposes of vehicle travel time prediction. It suggests that, add travel time observations from the previous departure time intervals are beneficial to the prediction, particularly when there is no other types of real-time information (e.g. traffic flow, speed) are available. It was also found that modular model fitting does not improve the quality of the prediction in all experimental settings used in this paper. Research limitations/implications The findings are primarily based on empirical studies on limited real-life data instances, and the results may lack generalisabilities. Therefore, the researchers are encouraged to test them further in more real-life data instances. Practical implications The paper includes implications and guidelines for the development of efficient GPS data storage and high-quality travel time prediction under different types of travel conditions. Originality/value This paper systematically studies the combinatorial feature effects for tree-ensemble-based travel time prediction approaches.
... Travel time consists of running time between stations and dwell time at stations. Studies show that running time between stations is related with traffic condition and traffic signal [18], [21], [22]. Dwell time contains the time spent on opening the door, passengers' boarding and alighting, closing the door, which consume up to 26% of total travel time [23]. ...
Full-text available
Article
The operational stability of public transport is significant for both passengers and operators. Affected by many stochastic factors such as traffic congestion, traffic signals and passenger demand at stops, the headway always become uneven, which greatly reduces the service quality. This paper used big global positioning systems (GPS) trajectory data to analyze the headway stability of bus system from the perspective of network. Statistical method is proposed to analyze the operational vehicle performance of bus network. GPS trajectory data of Jinan is used to test the model. The results show that the average dwell time, actual headway and headway stability index of stations follow lognormal distributions with obvious right tails. Moreover, the seriously unstable situations do not appear in the peak hours, but in the time periods before peak hours. In addition, the stations with most unstable headway are located in the suburbs and the fringe area of downtown. The outcomes suggest that operators should pay more attention to the suburbs and the fringe area of downtown and the time periods before peak hours to efficiently improve the service quality.
Article
The increase in population and the crowding of cities bring along transportation problems. Thus, people are directed to public transportation to reduce the burden on transportation. Being informed correctly about the arrival time at the stops attracts passengers. In this study, machine learning methods with three-layer architecture were used to predict bus arrival time. The first layer processes the measured data and give the prediction results of actual data. In the second layer, the residuals are predicted at the specified depth. In the third layer, the results of the previous two layers are integrated with three different approaches to calculate the final prediction. The case study was carried out on the data obtained from Istanbul public transportation and various machine learning methods were applied on the data using the traditional and the three-layer architecture. The experimental results showed that the three-layer architecture provided successful results with approximately 2.552 MAPE.
Full-text available
Chapter
This paper investigates the application of a set of machine learning algorithms to predict the time that will be spent inside a vehicle between any two locations. The ride duration is estimated by analyzing data collected from historical traces of taxis, i.e., Jan 2015 New York City (NYC) Yellow Cab trip record data. Moreover, the taxi data is integrated with Uber dataset to estimate time accurately taking into account a set of semantic variables. Nevertheless, these semantic variables are selected through outlier detection and feature selection using Chi-Square scores. Features such as pick-up latitude and longitude, drop-off latitude and longitude, pick-up date, pick-up time, etc., are considered for prediction purposes in NYC dataset. Mainly, the forecasting effectiveness is compared for three machine learning models, namely, Decision Tree Regression (DTR), Random Forest Regression (RFR), and K-Nearest Neighbor Regression (KNNR). It is found that RFR and KNNR are the favorites for this travel time prediction. Also, our result supersedes the best performance of Kaggle competition.
Article
The acquisition of massive data on parcel delivery motivates postal operators to foster the development of predictive systems to improve customer service. Predicting delivery times successive to being shipped out of the final depot, referred to as last-mile prediction, deals with complicating factors such as traffic, drivers’ behaviors, and weather. This work studies the use of deep learning for solving a real-world case of last-mile parcel delivery time prediction. We present our solution under the Internet-of-Things (IoT) paradigm and discuss its feasibility on a cloud-based architecture as a smart city application. We focus on a large-scale parcel data set provided by Canada Post, covering the Greater Toronto Area (GTA). We utilize an origin-destination (OD) formulation, in which routes are not available, but only the start and end delivery points. We investigate three categories of convolutional-based neural networks and assess their performances on the task. We further demonstrate how our modeling outperforms several baselines, from classical machine learning models to referenced OD solutions. We perform a thorough error analysis across the data and visualize the deep features learned to better understand the model behavior, making interesting remarks on data predictability. Our work provides an end-to-end neural pipeline that leverages parcel OD data as well as weather to accurately predict delivery durations. We believe that our system has the potential not only to improve user experience by better modeling their anticipation but also to aid last-mile postal logistics as a whole.
Full-text available
Chapter
The field of process mining focuses on the analysis of event data, generated and captured during the execution of processes within companies. The majority of existing process mining techniques focuses on process discovery, i.e., automated (data-driven) discovery of a descriptive process model of the process, and conformance and/or compliance checking. However, to effectively improve processes, a detailed understanding in differences of the actual performance of a process, as well as the underlying causing factors, is needed. Surprisingly, few research focuses on generic techniques for process-aware data-driven performance measurement, analysis and prediction. Therefore, in this paper, we present a generic approach, which allows us to compute the average performance between arbitrary groups of activities active in a process. In particular, the technique requires no a priori knowledge of the process, and thus does not suffer from representational bias induced by any underlying process representation. Our experiments show that our approach is scalable to large cases and especially robust to recurrent activities in a case.
Article
In this paper, we address the problem of travel time prediction of bus journeys which consist of bus riding times (may involve multiple bus services) and also the waiting times at transfer points. We propose a novel method called Traffic Pattern centric Segment Coalescing Framework (TP-SCF) that relies on learned disparate patterns of traffic conditions across different bus line segments for bus journey travel time prediction. Specifically, the proposed method consists of a training and a prediction stage. In the training stage, the bus lines are partitioned into bus line segments and the common travel time patterns of segments from different bus lines are explored using Non-negative Matrix Factorization (NMF). Bus line segments with similar patterns are classified into the same cluster. The clusters are then coalesced in order to extract data records for model training and bus journey time prediction. A separate Long Short Term Memory (LSTM) based model is trained for each cluster to predict the bus travel time under various traffic conditions. During prediction, a given bus journey is partitioned into the riding time components and waiting time components. The riding time components are predicted using the corresponding LSTM models of the clusters while the waiting time components are estimated based on historical bus arrival time records. We evaluated our method on large scale real-world bus travel data involving 30 bus services, and the results show that the proposed method notably outperforms the state-of-the-art approaches for all the scenarios considered.
Article
This paper presents a data-driven hybrid control (DDHC) framework that can arrange adaptive control strategies for vehicles to effectively improve the transit performance of the public transport system. The framework depicts a powerful combination of a data-driven control method that is used to imitate the control behaviour of dispatchers and a mathematical optimization method. Three components comprise the DDHC framework: a data-driven control module, a performance module, and an optimization module. The data-driven control module contains a random forest model which is adopted to justify whether to intervene in the operation of a bus line, and if so, which vehicles should be controlled and what type of control strategy should be taken – an acceleration strategy or deceleration strategy. The performance module including vehicle operation state models is used to describe the system evolution. The last component optimizes the specific control actions – which type of acceleration or deceleration strategy should be adopted – by minimizing total passenger travel time. The effectiveness of the proposed DDHC framework is evaluated with the data of a transit route in Urumqi, China. The results show that the DDHC framework with reasonable parameters can suit the needs of real-time control in complex traffic environments.
Chapter
Travel time prediction is an important issue for many people who want to know their departure time from an origin and arrival time at a destination in order to make decisions (e.g., postpone departure time at certain hours) and to reduce their waiting time at bus stops. Accurate predictions of bus travel time are necessary to know whether the travel time over target intervals between pairs of adjacent bus stops is stable or not. For this purpose, at first, we classified intervals between adjacent bus stops into two classes: stable and unstable. Next, we identified two statistically significant factors: variations of travel time in the same time periods over days and correlation of travel time between eight time-periods, which influence the bus travel time in the current time-period over unstable intervals. Then, we developed nonlinear dynamical models for predicting bus travel time over each unstable interval between adjacent bus stops for 7 time periods in a day. The proposed method basically utilizes time series methods based on Artificial Neural Network (ANN), Support Vector Machine Regression (SVR) and Random Forest (RF). We conducted experiments using bus probe data collected from November 21st to December 20th, 2013 and provided by Nishitetsu Bus Company, Fukuoka, Japan. In addition, to evaluate our proposed approach, we conducted a comparison experiment between our proposed model and the model in our previous study. Experimental results show that our proposed models can effectively improve the previous study model on the prediction accuracy of travel times over unstable intervals.
Full-text available
Article
Motivated by interest in making delay announcements to arriving customers who must wait in call centers and related service systems, we study the performance of alternative real-time delay estimators based on recent customer delay experience. The main estimators considered are: (i) the delay of the last customer to enter service (LES), (ii) the delay experienced so far by the customer at the head of the line (HOL), and (iii) the delay experienced by the customer to have arrived most recently among those who have already completed service (RCS). We compare these delay-history estimators to the standard estimator based on the queue length (QL), commonly used in practice, which requires knowledge of the mean interval between successive service completions in addition to the queue length. We characterize performance by the mean squared error (MSE). We do analysis and conduct simulations for the standard GI=M=s multi- server queueing model, emphasizing the case of large s. We obtain analytical results for the conditional distribution of the delay given the observed HOL delay. An approximation to its mean value serves as a reflned estimator. For all three candidate delay estimators, the MSE relative to the square of the mean is asymptotically negligible in the many-server and classical heavy-tra-c limiting regimes.
Full-text available
Article
This paper proposes a new tree-based ensemble method for supervised classification and regression problems. It essentially consists of randomizing strongly both attribute and cut-point choice while splitting a tree node. In the extreme case, it builds totally randomized trees whose structures are independent of the output values of the learning sample. The strength of the randomization can be tuned to problem specifics by the appropriate choice of a parameter. We evaluate the robustness of the default choice of this parameter, and we also provide insight on how to adjust it in particular situations. Besides accuracy, the main strength of the resulting algorithm is computational efficiency. A bias/variance analysis of the Extra-Trees algorithm is also provided as well as a geometrical and a kernel characterization of the models induced.
Full-text available
Article
Scikit-learn is a Python module integrating a wide range of state-of-the-art machine learning algorithms for medium-scale supervised and unsupervised problems. This package focuses on bringing machine learning to non-specialists using a general-purpose high-level language. Emphasis is put on ease of use, performance, documentation, and API consistency. It has minimal dependencies and is distributed under the simplified BSD license, encouraging its use in both academic and commercial settings. Source code, binaries, and documentation can be downloaded from http://scikit-learn.sourceforge.net.
Full-text available
Article
Modern Internet services, such as those at Google, Yahoo!, and Amazon, handle billions of requests per day on clusters of thousands of computers. Because these services operate under strict performance requirements, a statistical understanding of their performance is of great practical interest. Such services are modeled by networks of queues, where each queue models one of the computers in the system. A key challenge is that the data are incomplete, because recording detailed information about every request to a heavily used system can require unacceptable overhead. In this paper we develop a Bayesian perspective on queueing models in which the arrival and departure times that are not observed are treated as latent variables. Underlying this viewpoint is the observation that a queueing model defines a deterministic transformation between the data and a set of independent variables called the service times. With this viewpoint in hand, we sample from the posterior distribution over missing data and model parameters using Markov chain Monte Carlo. We evaluate our framework on data from a benchmark Web application. We also present a simple technique for selection among nested queueing models. We are unaware of any previous work that considers inference in networks of queues in the presence of missing data.
Full-text available
Article
In this paper, an overview of different analytic queueing models for traffic on road networks is presented. In the literature, it has been shown that queueing models can be used to adequately model uninterrupted traffic flows. This paper gives a broad review on this literature. Moreover, it is shown that the developed published methodologies (which are mainly single node oriented) can be extended towards queueing networks. First, an extension towards queueing networks with infinite buffer sizes is evaluated. Secondly, the assumption of infinite buffer sizes is dropped leading to queueing networks with finite buffer sizes. The impact of the buffer size when comparing the different queueing network methodologies is studied in detail. The paper ends with an analytical application tool to facilitate the optimal positioning of the counting points on a highway.
Full-text available
Article
Travel time is a fundamental measure in transportation. Accurate travel-time prediction also is crucial to the development of intelligent transportation systems and advanced traveler information systems. We apply support vector regression (SVR) for travel-time prediction and compare its results to other baseline travel-time prediction methods using real highway traffic data. Since support vector machines have greater generalization ability and guarantee global minima for given training data, it is believed that SVR will perform well for time series analysis. Compared to other baseline predictors, our results show that the SVR predictor can significantly reduce both relative mean errors and root-mean-squared errors of predicted travel times. We demonstrate the feasibility of applying SVR in travel-time prediction and prove that SVR is applicable and performs well for traffic data analysis.
Conference Paper
Information systems have been widely adopted to support service processes in various domains, e.g., in the telecommunication, finance, and health sectors. Recently, work on process mining showed how management of these processes, and engineering of supporting systems, can be guided by models extracted from the event logs that are recorded during process operation. In this work, we establish a queueing perspective in operational process mining. We propose to consider queues as first-class citizens and use queueing theory as a basis for queue mining techniques. To demonstrate the value of queue mining, we revisit the specific operational problem of online delay prediction: using event data, we show that queue mining yields accurate online predictions of case delay.
Conference Paper
In service processes, as found in the telecommunications, financial, or healthcare sector, customers compete for the scarce capacity of service providers. For such processes, performance analysis is important and it often targets the time that customers are delayed prior to service. However, this wait time cannot be fully explained by the load imposed on service providers. Indeed, it also depends on resource scheduling protocols, which determine the order of activities that a service provider decides to follow when serving customers. This work focuses on automatically learning resource decisions from events. We hypothesize that queueing information serves as an essential element in mining such protocols and hence, we utilize the queueing perspective of customers in the mining process. We propose two types of mining techniques: advanced classification methods from data mining that include queueing information in their explanatory features and heuristics that originate in queueing theory. Empirical evaluation shows that incorporating the queueing perspective into mining of scheduling protocols improves predictive power.
Article
Function estimation/approximation is viewed from the perspective of numerical optimization iti function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest-descent minimization. A general gradient descent "boosting" paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are regression trees, and tools for interpreting such "TreeBoost" models are presented. Gradient boosting of regression trees produces competitives highly robust, interpretable procedures for both regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire and Friedman, Hastie and Tibshirani are discussed.
Article
Bagging predictors is a method for generating multiple versions of a predictor and using these to get an aggregated predictor. The aggregation averages over the versions when predicting a numerical outcome and does a plurality vote when predicting a class. The multiple versions are formed by making bootstrap replicates of the learning set and using these as new learning sets. Tests on real and simulated data sets using classification and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy. The vital element is the instability of the prediction method. If perturbing the learning set can cause significant changes in the predictor constructed, then bagging can improve accuracy.
Conference Paper
Public transit, specially the bus, are usually considered unreliable as arrival times do not often match with the scheduled times. Current journey planners do not take this uncertainty into account, for example do not account for the risk of missing a transfer and of delays due to traffic conditions. This paper deals with analytics for building a multi modal journey planner, by estimating arrival and travel times for a certain trip involving transfers, and taking into account the delay of buses and risk of missing connections. We develop a system to compute probability distributions of end-to-end travel times, taking into account uncertain departure, travel and transfer times. We tested the system using real historical data for the whole bus system in Dublin, and showed the uncertainty of scheduled travel times with respect to the stochasticity of real traffic conditions. This is a first step towards building an uncertainty-aware multi-modal journey planner.
Conference Paper
Predicting arrival times of buses is a key challenge in the context of building intelligent public transportation systems. In this paper, we describe an efficient non-parametric algorithm which provides highly accurate predictions based on real-time GPS measurements. The key idea is to use a Kernel Regression model to represent the dependencies between position updates and the arrival times at bus stops. The performance of the proposed algorithm is evaluated on real data from the public bus transportation system in Dublin, Ireland. For a time horizon of 50 minutes, the prediction error of the algorithm is less than 10 percent on average. It clearly outperforms parametric methods which use a Linear Regression model, predictions based on the K-Nearest Neighbor algorithm, and a system which computes predictions of arrival times based on the current delay of buses. A study investigating the selection of interpolation points to reduce the size of the training set concludes the paper.
Article
Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, ***, 148–156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.
Article
Effective prediction of bus arrival time is central to many advanced traveler information systems. This article presents support vector machines (SVM), a new neural network algorithm, to predict bus arrival time. The objective of this paper is to examine the feasibility and applicability of SVM in vehicle travel time forecasting area. Segment, the travel time of current segment, and the latest travel time of next segment are taken as three input features. Bus arrival time predicted by the SVM is assessed with the data of transit route number 4 in Dalian economic and technological development zone in China and conclusions are drawn.
Article
Transit operations are interrupted frequently by stochastic variations in traffic and ridership conditions that deteriorate schedule or headway adherence and thus lengthen passenger wait times. Providing passengers with accurate vehicle arrival information through advanced traveler information systems is vital to reducing wait time. Two artificial neural networks (ANNs), trained by link-based and stop-based data, are applied to predict transit arrival times. To improve prediction accuracy, both are integrated with an adaptive algorithm to adapt to the prediction error in real time. The bus arrival times predicted by the ANNs are assessed with the microscopic simulation model CORSIM, which has been calibrated and validated with real-world data collected from route number 39 of the New Jersey Transit Corporation. Results show that the enhanced ANNs outperform the ones without integration of the adaptive algorithm.
Article
This study investigates the effects of adding anionic and nonionic surfactants to the scrubbing liquid during the absorption of naphthalene (Nap) using a wet scrubber. Both batch and continuous experiments were performed on a laboratory-scale packed tower scrubber by adding surfactants to the scrubbing liquid. An anionic surfactant sodium dodecyl sulfate (SDS) and two nonionic polyoxyethylene surfactants C10E4 [tetraethylene glycol mono(decyl ether)] and C14E8 [octaethylene glycol mono(tetradecyl ether)] were used. The concentrations of the surfactant solutions all exceeded the critical micelle concentration. The results of continuous experiments indicate that the efficiencies of removal of naphthalene by C10E4 and C14E8 were 75.0 and 71.9%, respectively, at a concentration of 1.0×10-2 M. The removal efficiency of naphthalene by SDS ranged from 6 to 39% at concentrations from 1.0×10-2 to 1.0×10-1 M under continuous scrubbing. In the batch experiments, the batch naphthalene absorption capacities were estimated to be 31.8 μg Nap/g C10E4, 12.9 μg Nap/g C14E8, and 2.4 μg Nap/g SDS. A comparison was also made among SDS, C10E4, and C14E8 in terms of cost, foaming, impact of wastewater, removal efficiency, and absorption capacity.
Article
The objective of this research is to develop a dynamic model to forecast multi-interval path travel times between bus stops of origin and destination. The research also intends to test the proposed model using real-world data. This research was brought about by the shortcomings of the existing real-time based short-term-prediction models, which have been widely utilised for single interval predictions. The developed model is based on the Nearest Neighbour Non-Parametric Regression using historical and current data collected by the Automatic Vehicle Location technology. In a test with real-world bus data in Seoul, Korea, the proposed multi-interval-prediction model performed effectively in terms of both prediction accuracy and computing time.
Article
Bagging predictors is a method for generating multiple versions of a predictor and using these to get an aggregated predictor. The aggregation averages over the versions when predicting a numerical outcome and does a plurality vote when predicting a class. The multiple versions are formed by making bootstrap replicates of the learning set and using these as new learning sets. Tests on real and simulated data sets using classification and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy. The vital element is the instability of the prediction method. If perturbing the learning set can cause significant changes in the predictor constructed, then bagging can improve accuracy.
Article
In this paper we consider an open queueing network having multiple classes, priorities, and general service time distributions. In the case where there is a single bottleneck station we conjecture that normalized queue length and sojourn time processes converge, in the heavy traffic limit, to one-dimensional reflected Brownian motion, and present expressions for its drift and variance. The conjecture is motivated by known heavy traffic limit theorems for some special cases of the general model, and some conjectured “Heavy Traffic Principles” derived from them. Using the known stationary distribution of one-dimensional reflected Brownian motion, we present expressions for the heavy traffic limit of stationary queue length and sojourn time distributions and moments. For systems with Markov routing we are able to explicitly calculate the limits.
Article
Function approximation is viewed from the perspective of numerical optimization in function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest--descent minimization. A general gradient--descent "boosting" paradigm is developed for additive expansions based on any fitting criterion. Specific algorithms are presented for least--squares, least--absolute--deviation, and Huber--M loss functions for regression, and multi--class logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are decision trees, and tools for interpreting such "TreeBoost" models are presented. Gradient boosting of decision trees produces competitive, highly robust, interpretable procedures for regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire 1996, and Fr...