ChapterPDF Available

A Comparative Study of Big Mart Sales Prediction

Authors:

Abstract and Figures

Nowadays shopping malls and Big Marts keep the track of their sales data of each and every individual item for predicting future demand of the customer and update the inventory management as well. These data stores basically contain a large number of customer data and individual item attributes in a data warehouse. Further, anomalies and frequent patterns are detected by mining the data store from the data warehouse. The resultant data can be used for predicting future sales volume with the help of different machine learning techniques for the retailers like Big Mart. In this paper, we propose a predictive model using Xgboost technique for predicting the sales of a company like Big Mart and found that the model produces better performance as compared to existing models. A comparative analysis of the model with others in terms performance metrics is also explained in details.
Content may be subject to copyright.
See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/336530068
A Comparative Study of Big Mart Sales Prediction
Conference Paper · September 2019
CITATION
1
READS
7,893
2 authors:
Gopal Behera
Government College of Engineering Kalahandi,Bhawanipatna,Odisha,India
7 PUBLICATIONS11 CITATIONS
SEE PROFILE
Neeta Nain
Malaviya National Institute of Technology Jaipur
7 PUBLICATIONS1 CITATION
SEE PROFILE
All content following this page was uploaded by Gopal Behera on 23 October 2019.
The user has requested enhancement of the downloaded file.
A Comparative Study of Big Mart Sales
Prediction
Gopal Behera1and Neeta Nain2
Malaviya National Institute of Technology Jaipur, India1,2
2019rcp9002@mnit.ac.in1,nnain.cse@mnit.ac.in2
Abstract. Nowadays shopping malls and Big Marts keep the track of
their sales data of each and every individual item for predicting future
demand of the customer and update the inventory management as well.
These data stores basically contain a large number of customer data and
individual item attributes in a data warehouse. Further, anomalies and
frequent patterns are detected by mining the data store from the data
warehouse. The resultant data can be used for predicting future sales
volume with the help of different machine learning techniques for the
retailers like Big Mart. In this paper, we propose a predictive model using
Xgboost technique for predicting the sales of a company like Big Mart
and found that the model produces better performance as compared to
existing models. A comparative analysis of the model with others in
terms performance metrics is also explained in details.
Keywords: Machine Learning, Sales Forecasting, Random Forest, Re-
gression, Xgboost.
1 INTRODUCTION
Day by day competition among different shopping malls as well as big marts is
getting more serious and aggressive only due to the rapid growth of the global
malls and on-line shopping. Every mall or mart is trying to provide personal-
ized and short-time offers for attracting more customers depending upon the
day, such that the volume of sales for each item can be predicted for inventory
management of the organization, logistics and transport service, etc. Present
machine learning algorithm are very sophisticated and provide techniques to
predict or forecast the future demand of sales for an organization, which also
helps in overcoming the cheap availability of computing and storage systems.
In this paper, we are addressing the problem of big mart sales prediction or
forecasting of an item on customer’s future demand in different big mart stores
across various locations and products based on the previous record. Different
machine learning algorithms like linear regression analysis, random forest, etc
are used for prediction or forecasting of sales volume. As good sales are the life
of every organization so the forecasting of sales plays an important role in any
shopping complex. Always a better prediction is helpful, to develop as well as to
enhance the strategies of business about the marketplace which is also helpful
2 Gopal Behera and Neeta Nain
to improve the knowledge of marketplace. A standard sales prediction study can
help in deeply analyzing the situations or the conditions previously occurred and
then, the inference can be applied about customer acquisition, funds inadequacy
and strengths before setting a budget and marketing plans for the upcoming
year. In other words, sales prediction is based on the available resources from
the past. In depth knowledge of past is required for enhancing and improving
the likelihood of marketplace irrespective of any circumstances especially the
external circumstance, which allows to prepare the upcoming needs for the busi-
ness. Extensive research is going on in retailers domain for forecasting the future
sales demand. The basic and foremost technique used in predicting sale is the
statistical methods, which is also known as the traditional method, but these
methods take much more time for predicting a sales also these methods could
not handle non linear data so to over these problems in traditional methods
machine learning techniques are deployed. Machine learning techniques can not
only handle non-linear data but also huge data-set efficiently. To measure the
performance of the models, Root Mean Square Error (RMSE) [15] and Mean
Absolute Error (MAE) [4] are used as an evaluation metric as mentioned in the
Equation 1 and 2 respectively. Here Both metrics are used as the parameter for
accuracy measure of a continuous variable.
MAE =1
n
n
X
i=1
xpredict xactual
(1)
RM SE =v
u
u
t
1
n
n
X
i=1
xpredict xactual
2(2)
where n: total number of error and
xpredict xactual
: Absolute error. The
remaining part of this article is arranged as following: Section 1 briefly describes
introduction of sales prediction of Big Mart and also elaborate about the evalua-
tion metric used in the model. Previous related work has been pointed in Section
2. The detailed description and analysis of proposed model is given in Section
3. Where as implementations and results are demonstrated in Section 4 and the
paper concludes with a conclusion in the last section.
2 Related Work
Sales forecasting as well as analysis of sale forecasting has been conducted by
many authors as summarized: The statistical and computational methods are
studied in [2] also this paper elaborates the automated process of knowledge
acquisition. Machine learning [6] is the process where a machine will learn from
data in the form of statistically or computationally method and process knowl-
edge acquisition from experiences. Various machine learning (ML) techniques
with their applications in different sectors has been presented in [2]. Pat Lang-
ley and Herbert A [7] pointed out most widely used data mining technique in
A Comparative Study of Big Mart Sales Prediction 3
the field of business is the Rule Induction (RI) technique as compared to other
data mining techniques. Where as sale prediction of a pharmaceutical distri-
bution company has been described in [12,10]. Also this paper focuses on two
issues: (i) stock state should not undergo out of stock, and (ii) it avoids the
customer dissatisfaction by predicting the sales that manages the stock level of
medicines. Handling of footwear sale fluctuation in a period of time has been
addressed in [5]. Also this paper focuses on using neural network for predicting
of weekly retail sales, which decrease the uncertainty present in the short term
planning of sales. Linear and non-linear [3] a comparative analysis model for sales
forecasting is proposed for the retailing sector. Beheshti-Kashi and Samaneh [1]
performed sales prediction in fashion market. A two level statistical method [11]
is elaborated for forecasting the big mart sales prediction. Xia and Wong [17]
proposed the differences between classical methods (based on mathematical and
statistical models) and modern heuristic methods and also named exponential
smoothing, regression, auto regressive integrated moving average (ARIMA), gen-
eralized auto regressive conditionally heteroskedastic (GARCH) methods. Most
of these models are linear and are not able to deal with the asymmetric behavior
in most real-world sales data [9]. Some of the challenging factors like lack of
historical data, consumer-oriented markets face uncertain demands, and short
life cycles of prediction methods results in inaccurate forecast.
3 Proposed System
For building a model to predict accurate results the dataset of Big Mart sales
undergoes several sequence of steps as mentioned in Figure 1 and in this work we
propose a model using Xgboost technique. Every step plays a vital role for build-
ing the proposed model. In our model we have used 2013 Big mart dataset [13].
After preprocessing and filling missing values, we used ensemble classifier using
Decision trees, Linear regression, Ridge regression, Random forest and Xgboost.
Both MAE and RSME are used as accuracy metrics for predicting the sales in
Big Mart. From the accuracy metrics it was found that the model will predict
best using minimum MAE and RSME. The details of the proposed method is
explained in the following section.
3.1 Dataset Description of Big Mart
In our work we have used 2013 Sales data of Big Mart as the dataset. Where
the dataset consists of 12 attributes like Item Fat, Item Type, Item MRP,
Outlet Type, Item Visibility, Item Weight, Outlet Identifier, Outlet Size, Outlet
Establishment Year, Outlet Location Type, Item Identifier and Item Outlet Sales.
Out of these attributes response variable is the Item Outlet Sales attribute and
remaining attributes are used as the predictor variables. The data-set consists of
8523 products across different cities and locations. The data-set is also based on
hypotheses of store level and product level. Where store level involves attributes
like: city, population density, store capacity, location, etc and the product level
4 Gopal Behera and Neeta Nain
Fig. 1. Working procedure of proposed model.
hypotheses involves attributes like: brand, advertisement, promotional offer, etc.
After considering all, a dataset is formed and finally the data-set was divided
into two parts, training set and test set in the ratio 80 : 20.
3.2 Data Exploration
In this phase useful information about the data has been extracted from the
dataset. That is trying to identify the information from hypotheses vs available
data. Which shows that the attributes Outlet size and Item weight face the
problem of missing values, also the minimum value of Item Visibility is zero
which is not actually practically possible. Establishment year of Outlet varies
from 1985 to 2009. These values may not be appropriate in this form. So, we
need to convert them into how old a particular outlet is. There are 1559 unique
products, as well as 10 unique outlets, present in the dataset. The attribute
Item type contains 16 unique values. Where as two types of Item Fat Content
are there but some of them are misspelled as regular instead of ’Regular’ and
low fat, LF instead of Low Fat. From Figure 2. It was found that the response
variable i.e. Item Outlet Sales was positively skewed. So, to remove the skewness
of response variable a log operation was performed on Item Outlet Sales.
3.3 Data Cleaning
It was observed from the previous section that the attributes Outlet Size and
Item Weight has missing values. In our work in case of Outlet Size missing
A Comparative Study of Big Mart Sales Prediction 5
Fig. 2. Univariate distribution of target variable Item outlet sales. The Target variable
is positively skewed towards the higher sales.
value we replace it by the mode of that attribute and for the Item Weight
missing values we replace by mean of that particular attribute. The missing
attributes are numerical where the replacement by mean and mode diminishes
the correlation among imputed attributes. For our model we are assuming that
there is no relationship between the measured attribute and imputed attribute.
3.4 Feature Engineering
Some nuances were observed in the data-set during data exploration phase. So
this phase is used in resolving all nuances found from the dataset and make them
ready for building the appropriate model. During this phase it was noticed that
the Item visibility attribute had a zero value, practically which has no sense.
So the mean value item visibility of that product will be used for zero values
attribute. This makes all products likely to sell. All categorical attributes dis-
crepancies are resolved by modifying all categorical attributes into appropriate
ones. In some cases, it was noticed that non-consumables and fat content prop-
erty are not specified. To avoid this we create a third category of Item fat content
i.e. none. In the Item Identifier attribute, it was found that the unique ID starts
with either DR or FD or NC. So, we create a new attribute Item Type New with
three categories like Foods, Drinks and Non-consumables. Finally, for determin-
ing how old a particular outlet is, we add an additional attribute Year to the
dataset.
3.5 Model Building
After completing the previous phases, the dataset is now ready to build proposed
model. Once the model is build it is used as predictive model to forecast sales
of Big Mart. In our work, we propose a model using Xgboost algorithm and
compare it with other machine learning techniques like Linear regression, Ridge
regression [14], Decision tree [8,16] etc.
Decision Tree: A decision tree classification is used in binary classification prob-
lem and it uses entropy [8] and information gain [16] as metric and is defined in
6 Gopal Behera and Neeta Nain
Equation 3 and Equation 4 respectively for classifying an attribute which picks
the highest information gain attribute to split the data set.
H(S) = X
cC
p(c) log p(c) (3)
where H(S): Entropy, C: Class Label, P:Probability of class c.
Infromation Gain(S, A) = H(S)X
tT
p(t)H(t) (4)
where S: Set of attribute or dataset, H(S): Entropy of set S,T: Subset created
from splitting of Sby attribute A.p(t): Proportion of the number of elements
in tto number of element in the set S.H(t): Entropy of subset t. The decision
tree algorithm is depicted in Algorithm 1.
Require: Set of features dand set of training instances D
1: if all the instances in Dhave the same target label Cthen
2: Return a decision tree consisting of leaf node with label level C
end
else if dis empty then
4: Return a decision tree of leaf node with label of the majority
target level in D
end
5: else if Dis empty then
6: Return a decision tree of leaf node with label of the majority
target level of the immediate parent node
end
7: else
8: d[best]arg max IG(d, D) where dD
9: make a new node, Noded[best]
10: partition Dusing d[best]
11: remove d[best] from d
12: for each partition Diof Ddo
13: grow a branch from Noded[best]to the decision tree created by
rerunning ID3 with D=Di
end
end
Algorithm 1: I D3 algorithm
Linear Regression: A model which create a linear relationship between the
dependent variable and one or more independent variable, mathematically linear
regression is defined in Equation 5
y=wTx(5)
where yis dependent variable and xare independent variables or attributes. In
linear regression we find the value of optimal hyperplane wwhich corresponds
to the best fitting line (trend) with minimum error. The loss function for linear
A Comparative Study of Big Mart Sales Prediction 7
regression is estimated in terms of RMSE and MAE as mentioned in the Equa-
tion 1 and 2.
Ridge Regression: The cost function for ridge regression is defined in Equation 6.
min |(YX(θ)|)2+λkθk2(6)
here λknown as the penalty term as denoted by αparameter in the ridge func-
tion. So the penalty term is controlled by changing the values of α, higher the
values of αbigger is the penalty. Figure 3 shows Linear Regression, Ridge Re-
gression, Decision Tree and proposed model i.e. Xgboost.
Xgboost (Extreme Gradient Boosting) is a modified version of Gradient Boosting
Machines (GBM) which improves the performance upon the GBM framework
by optimizing the system using a differentiable loss function as defined in Equa-
tion 7. n
X
i=1
l(yi,ˆyi) + X
k
kΩ(fk), fkF(7)
where ˆyi: is the predicted value, yi: is the actual value and Fis the set of
function containing the tree, l(yi,ˆyi) is the loss function.
This enhances the GBM algorithm so that it can work with any differentiable
loss function. The GBM algorithm is illustrated in Algorithm 2.
Step 1: Initialize model with a constant value:
F0=arg min
n
X
i=0
L(yi, γ)
Step 2: for m= 1 to M : do
a. Compute pseudo residuals:
rim =∂L(yiF(xi))
∂F (xi)F(x)=Fm1(x)
for all i = 1,2...n
b. Fit a Base learner hm(x) to pseudo residuals that is train the
learner using training set.
c. Compute γm
γm=γarg min
n
X
i=0
(L(yi, Fm1(xi) + γh(xi)))
d. Update the model:
Fm(x) = Fm1(x) + γmhm(x)
end
Step 3: Output FM
Algorithm 2: Gradient boosting machine(GBM) algorithm
8 Gopal Behera and Neeta Nain
The Xgboost has following exclusive features:
1. Sparse Aware - that is the missing data values are automatic handled.
2. Supports parallelism of tree construction.
3. Continued training - so that the fitted model can further boost with new
data.
All models received features as input, which are then segregated into training
and test set. The test dataset is used for sales prediction.
Fig. 3. Framework of proposed model. Model received the input features and split it
into training and test set. The trained model is used to predict the future sales.
4 Implementation and Results
In our work we set cross-validation as 20 fold cross-validation to test accuracy
of different models. Where in the cross-validation stage the dataset is divided
randomly into 20 subsets with roughly equal sizes. Out of the 20 subsets, 19
subsets are used as training data and the remaining subset forms the test data
also called leave-one-out cross validation. Every models is first trained by using
the training data and then used to predict accuracy by using test data and this
continues until each subset is tested once.
From data visualization, it was observed that lowest sales were produced in
smallest locations. However, in some cases it was found that medium size loca-
tion produced highest sales though it was type-3 (there are three type of super
market e.g. super market type-1, type-2 and type-3) super market instead of
largest size location as shown in Figure 4. to increase the product sales of Big
mart in a particular outlet, more locations should be switched to Type 3 Super-
markets.
A Comparative Study of Big Mart Sales Prediction 9
Fig. 4. Impact of outlet location type on
target variable item outlet sale. Displayed
the sales volume of different outlet loca-
tions.
Fig. 5. Correlation among fea-
tures of a dataset. Brown
squares are highly correlated
whereas black square repre-
sents bad correlation among at-
tributes.
However, the proposed model gives better predictions among other models for
future sales at all locations. For example, how item MRP is correlated with outlet
sales is shown in Figure 5. Also Figure 5 shows that Item Outlet Sales is strongly
correlated with Item MRP, where the correlation is defined in Equation 8.
PCorr =nP(xy)(Px)(Py)
pn[Px2)(Px)2]n[Py2(Py)2](8)
From Figure 8 it is also observed that target attribute Item Outlet Sales is af-
fected by sales of the Item Type. Similarly, from Figure 6 it is also observed that
highest sales is made by OUT027 which is actually a medium size outlet in the
super market type-3. Figure 7 describes that the less visible products are sold
more compared to the higher visibility products which is not possible practically.
Thus, we should reject the null hypothesis H0- that the visibility does not effect
the sales.
From Figure 9 it is observed that less number of high outlet size stores
exist in comparison to the medium and small outlet size in terms of count.
The cross-validation score along with MAE and RMSE of the proposed model
and existing models is shown in Table 1 and Table 2 respectively. Similarly the
root mean squared error for existing model and proposed model is presented in
Table 2. From the results we observe that and found that the proposed model is
significantly improved over the other model.
5 Conclusions
In present era of digitally connected world every shopping mall desires to know
the customer demands beforehand to avoid the shortfall of sale items in all sea-
sons. Day to day the companies or the malls are predicting more accurately the
10 Gopal Behera and Neeta Nain
Fig. 6. Impact of outlet identifier
on target variable item outlet sale.
Fig. 7. Impact of item visibility
on target variable item outlet sale.
Less visible items are sold more
compared to more visibility items
as outlet contains daily used items
which contradicts the null hypoth-
esis.
Fig. 8. Impact of item type on tar-
get variable item outlet sale.
Fig. 9. Distribution of outlet size. The
number of outlet size are available in the
dataset.
Table 1. Comparison of Cross Validation Score of different model
Model Cross Validation
Score (Mean)
Cross Validation
Score(Std)
Linear Regression 1129 43.24
Decision Tree 1091 45.42
Ridge Regression 1097 43.41
Table 2. Comparison of MAE and RMSE of proposed model with other Model
Model MAE RMSE
Linear Regression 836.1 1127
Decision Tree 741.6 1058
Ridge Regression 836 1129
Xgboost 739.03 1052
A Comparative Study of Big Mart Sales Prediction 11
demand of product sales or user demands. Extensive research in this area at
enterprise level is happening for accurate sales prediction. As the profit made
by a company is directly proportional to the accurate predictions of sales, the
Big marts are desiring more accurate prediction algorithm so that the company
will not suffer any losses. In this research work, we have designed a predictive
model by modifying Gradient boosting machines as Xgboost technique and ex-
perimented it on the 2013 Big Mart dataset for predicting sales of the product
from a particular outlet. Experiments support that our technique produce more
accurate prediction compared to than other available techniques like decision
trees, ridge regression etc. Finally a comparison of different models is summa-
rized in Table 2. From Table 2 it is also concluded that our model with lowest
MAE and RMSE performs better compared to existing models.
References
1. Beheshti-Kashi, S., Karimi, H.R., Thoben, K.D., L¨utjen, M., Teucke, M.: A survey
on retail sales forecasting and prediction in fashion markets. Systems Science &
Control Engineering 3(1), 154–161 (2015)
2. Bose, I., Mahapatra, R.K.: Business data mininga machine learning perspective.
Information & management 39(3), 211–225 (2001)
3. Chu, C.W., Zhang, G.P.: A comparative study of linear and nonlinear models for
aggregate retail sales forecasting. International Journal of production economics
86(3), 217–231 (2003)
4. Claypool, M., Gokhale, A., Miranda, T., Murnikov, P., Netes, D., Sartin, M.:
Combing content-based and collaborative filters in an online newspaper (1999)
5. Das, P., Chaudhury, S.: Prediction of retail sales of footwear using feedforward and
recurrent neural networks. Neural Computing and Applications 16(4-5), 491–502
(2007)
6. Domingos, P.M.: A few useful things to know about machine learning. Commun.
acm 55(10), 78–87 (2012)
7. Langley, P., Simon, H.A.: Applications of machine learning and rule induction.
Communications of the ACM 38(11), 54–64 (1995)
8. Loh, W.Y.: Classification and regression trees. Wiley Interdisciplinary Reviews:
Data Mining and Knowledge Discovery 1(1), 14–23 (2011)
9. Makridakis, S., Wheelwright, S.C., Hyndman, R.J.: Forecasting methods and ap-
plications. John wiley & sons (2008)
10. Ni, Y., Fan, F.: A two-stage dynamic sales forecasting model for the fashion retail.
Expert Systems with Applications 38(3), 1529–1536 (2011)
11. Punam, K., Pamula, R., Jain, P.K.: A two-level statistical model for big mart sales
prediction. In: 2018 International Conference on Computing, Power and Commu-
nication Technologies (GUCON). pp. 617–620. IEEE (2018)
12. Ribeiro, A., Seruca, I., Dur˜ao, N.: Improving organizational decision support: De-
tection of outliers and sales prediction for a pharmaceutical distribution company.
Procedia Computer Science 121, 282–290 (2017)
13. Shrivas, T.: Big mart dataset@ONLINE (Jun 2013), https://datahack.
analyticsvidhya.com/contest/practice-problem-big-mart- sales-iii/
14. Smola, A.J., Sch¨olkopf, B.: A tutorial on support vector regression. Statistics and
computing 14(3), 199–222 (2004)
12 Gopal Behera and Neeta Nain
15. Smyth, B., Cotter, P.: Personalized electronic program guides for digital tv. Ai
Magazine 22(2), 89–89 (2001)
16. Wang, Y., Witten, I.H.: Induction of model trees for predicting continuous classes
(1996)
17. Xia, M., Wong, W.K.: A seasonal discrete grey forecasting model for fashion re-
tailing. Knowledge-Based Systems 57, 119–126 (2014)
View publication statsView publication stats
Article
Many online platforms have adopted a recommender system (RS) to suggest an actual product to the active users according to their preferences. The RS that provides accurate information on users’ past preferences is known as collaborative filtering (CF). One of the most common CF methods is matrix factorization (MF). It is important to note that the MF technique contains several tuned parameters, leading to an expensive and complex black-box optimization problem. An objective function quantifies the quality of a prediction by mapping any possible configuration of hyper-parameters to a numerical score. In this article, we show how a gird search optimization (GSO) can efficiently obtain the optimal value of hyper-parameters an MF and improve the prediction of the collaborative recommender system (CRS). Specifically, we designed a 4×4 grid search space, obtained the optimal set of hyper-parameters, and then evaluated the model using these hyper-parameters. Furthermore, we evaluated the model using two benchmark datasets and compared it with the state-of-the-art model. We found that the proposed model significantly improves the prediction accuracy, precision@k, and NDCG@k over the state-of-art-the models and handles the sparsity problem of CF.
Article
The essential concepts of sellers and buyers are supply and demand. Organizations must be able to accurately predict demand in order to establish plans. Sales Prediction is based on predicting sales for various Big Mart stores in order to adjust the business strategy based on the projected performance. For Big Mart firms, in this paper present a novel approach to demand forecast. The Big Mart firms' business model, for which the model is executed, contains multiple shops selling the same product at the same time across the country where the company maintains a marketplace model. This proposed Supervised and Artificial Neural Network Algorithms produces reliable results when compared to other learning methods. Feature selection, data transformation, and data exploration will all play essential roles. This method is used on data from Big-Mart Sales, where data is discovered, processed, and enough relevant data is taken to help forecast correct future outcomes.
Chapter
The advancements in deep learning methods have brought several new artificial intelligence (AI) applications making AI important for every enterprise that aims to be competitive. Therefore, not only Tech companies but also small- and medium-sized enterprises (SMEs) require AI. This paper discusses SME AI applications and reveals the challenges, solutions, and advantages of implementing AI in SMEs. Although some SMEs are concerned with building their applications because of the cost and length of implementing AI, resulting in a high risk of failure, nevertheless, SMEs still depend on artificial intelligence for growth and cloud-based solutions.KeywordsArtificial intelligenceDeep learningMachine learningSMEsEfficiencyDevelopmentApplications
Chapter
In today’s competitive market, every retail store needs to run its business successfully. Therefore they need to know the future demand of customers ahead. So that the retail store can have sufficient stock of their demanding product in their inventory. However, estimating future sales demand is an important and challenging task for any retailer store. In this paper, we use regression techniques to tackle such type of challenging task. Further, we investigate, evaluate, explore the merits and demerits of different regression-based forecasting techniques to predict weekly sales of a store. These forecasting techniques include linear, lasso, ridge, random forest, decision tree, extra tree, and XGBoost regression. Moreover, the effectiveness of the above models is measured through the performance metrics such as MAE, Weighted MAE (WMAE), and RMSE on a publicly available dataset. The experimental outcomes conclude that the random forest technique achieves better results than other methods.KeywordsRegressionKNNXGBoostSales forecastingDecision treeExtra treeAccuracy.
Article
In the modern era of reaching new lengths of advancement, every company and enterprise are working on their customer demands as well as their inventory management. The models used by them help them predict future demands by understanding the pattern from old sales records. Lately, everyone is abandoning the traditional prediction models for sales forecasting as it takes a prolonged amount of time to get the expected results. Therefore now the retailers keep track of their sales record in the form of a data set, which comprises price tag, outlet types, outlet location, item visibility, item outlet sales etc.
Chapter
Due to novel coronavirus (COVID-19), the world is facing a pandemic situation. Human lifestyle changed drastically during this pandemic period, and everyone is badly affected and do not know when the situation is going to be normal. Though the virus is under control, still there is a uncertainty and unpredictable situation exist not only in India but also all over the world. So it is very important to predict the COVID cases as early as possible so that the best precautionary measure can be taken. In this study, we have designed a parametric estimation curve using linear, exponential, and logistic model for forecasting new cases on 30 days ahead. From experimentation, it is found that the logistic model performs better than the linear and exponential model.KeywordsCOVID-19Parametric curveLinear modelExponential modelLogistic model
Article
The tremendous growth in information over the last decade leads to information overwhelming problems for accessing personalized products. The recommender framework that retrieves user preferences on past interactions is known as collaborative filtering (CF). Although, CF is a prevalent technique amongst the techniques applied in the recommender environment. However, it suffers from many problems like information sparsity, scalability, cold-start, etc. Many investigations have been made to tackle some of these issues with the help of matrix factorization (MF) approaches. However, MF cannot handle the nonlinearity among the data. Deep learning is an advanced learning technique that has shown success in many applications such as image classification, computer vision, natural language processing, etc. Little work has been reported on deep learning techniques in the recommender domain. We propose an efficient deep collaborative recommender system that embeds item metadata to handle the nonlinearity in data and sparsity. The model consists of two stages, wherein the first stage, a neural network, is used to retrieve the data’s nonlinear features through embedding vectors. These vectors are concatenated together and fed as input to the second stage of the model. The output of the model yields a partially observed rating. The input and the parameters are simultaneously optimized and updated to minimize errors. The proposed strategy is evaluated against the benchmark techniques on two well-known datasets. The exploratory outcomes signify our approach’s exactitude and efficiency. Moreover, the missing values can also be recovered by propagating the embedding vectors from the input to the output layers.
Conference Paper
Full-text available
Sales forecasting is an important aspect of different companies engaged in retailing, logistics, manufacturing, marketing and wholesaling. It allows companies to efficiently allocate resources, to estimate achievable sales revenue and to plan a better strategy for future growth of the company. In this paper, prediction of sales of a product from a particular outlet is performed via a two-level approach that produces better predictive performance compared to any of the popular single model predictive learning algorithms. The approach is performed on Big Mart Sales data of the year 2013. Data exploration, data transformation and feature engineering play a vital role in predicting accurate results. The result demonstrated that the two-level statistical approach performed better than a single model approach as the former provided more information that leads to better prediction.
Article
Full-text available
Stock unavailability in the supply of medicines to pharmacies can be caused by several factors including manufacturing problems, lack of raw materials, end of product selling, disease and epidemics outbreaks. Furthermore, the sale of medicines by some pharmacies to foreign markets has increased in recent years, and is considered one of the main causes of medicine supply failures in Portugal. This paper depicts the case study of a pharmaceutical distribution company in Portugal and aims to address two main research issues. The first one consisted in detecting customers (pharmacies) and products (medicines) which may be considered outliers and perform stock proration when these outliers are detected, in order to avoid abnormal sales and out-of-stocks in pharmacies. The second one targeted the sales prediction for the pharmaceutical distribution company, in order to better control and manage the levels of stock of medicines, so as to avoid excessive inventory costs while guaranteeing customer demand satisfaction, and thus decreasing the possibility of loss of customers due to stock outages. In outliers detection (customers and products) we used the Box-plot method as well as the SPSS statistical software. For sales prediction, the time series data mining method smoothed Pegels was used, while the implementation was done in SQL and the analyzed data was stored in an Oracle database.
Article
Full-text available
Sales forecasting is an essential task in retailing. In particular, consumer-oriented markets such as fashion and electronics face uncertain demands, short life cycles and a lack of historical sales data which strengthen the challenges of producing accurate forecasts. This survey paper presents state-of-the-art methods in the sales forecasting research with a focus on fashion and new product forecasting. This study also reviews different strategies to the predictive value of user-generated content and search queries.
Article
Full-text available
Classification and regression trees are machine-learning methods for constructing prediction models from data. The models are obtained by recursively partitioning the data space and fitting a simple prediction model within each partition. As a result, the partitioning can be represented graphically as a decision tree. Classification trees are designed for dependent variables that take a finite number of unordered values, with prediction error measured in terms of misclassification cost. Regression trees are for dependent variables that take continuous or ordered discrete values, with prediction error typically measured by the squared difference between the observed and predicted values. This article gives an introduction to the subject by reviewing some widely available algorithms and comparing their capabilities, strengths, and weakness in two examples. © 2011 John Wiley & Sons, Inc. WIREs Data Mining Knowl Discov 2011 1 14-23 DOI: 10.1002/widm.8
Article
MACHINE LEARNING SYSTEMS automatically learn programs from data. This is often a very attractive alternative to manually constructing them, and in the last decade the use of machine learning has spread rapidly throughout computer science and beyond. Machine learning is used in Web search, spam filters, recommender systems, ad placement, credit scoring, fraud detection, stock trading, drug design, and many other applications. A recent report from the McKinsey Global Institute asserts that machine learning (a.k.a. data mining or predictive analytics) will be the driver of the next big wave of innovation. 15 Several fine textbooks are available to interested practitioners and researchers (for example, Mitchell 16 and Witten et al. 24). However, much of the "folk knowledge" that is needed to successfully develop machine learning applications is not readily available in them. As a result, many machine learning projects take much longer than necessary or wind up producing less-than-ideal results. Yet much of this folk knowledge is fairly easy to communicate. This is the purpose of this article.
Article
In the fashion retail industry, level of forecasting accuracy plays a crucial role in retailers’ profit. In order to avoid stock-out and maintain a high inventory fill rate, fashion retailers require specific and accurate sales forecasting systems. One of the key factors of an effective forecasting system is the availability of long and comprehensive historical data. However, in the fashion retail industry, the sales data stored in the point-of-sales (POS) systems are always not comprehensive and scattered due to various reasons. This paper presents a new seasonal discrete grey forecasting model based on cycle truncation accumulation with amendable items to improve sales forecasting accuracy. The proposed forecasting model is to overcome two important problems: seasonality and limited data. Although there are several works suitable with one of them, there is no previous research effort that overcome both problems in the context of grey models. The proposed algorithms are validated using real POS data of three fashion retailers selling high-ended, medium and basic fashion items. It was found that the proposed model is practical for fashion retail sales forecasting with short historical data and outperforms other state-of-art forecasting techniques.
Article
The purpose of this paper is to compare the accuracy of various linear and nonlinear models for forecasting aggregate retail sales. Because of the strong seasonal fluctuations observed in the retail sales, several traditional seasonal forecasting methods such as the time series approach and the regression approach with seasonal dummy variables and trigonometric functions are employed. The nonlinear versions of these methods are implemented via neural networks that are generalized nonlinear functional approximators. Issues of seasonal time series modeling such as deseasonalization are also investigated. Using multiple cross-validation samples, we find that the nonlinear models are able to outperform their linear counterparts in out-of-sample forecasting, and prior seasonal adjustment of the data can significantly improve forecasting performance of the neural network model. The overall best model is the neural network built on deseasonalized time series data. While seasonal dummy variables can be useful in developing effective regression models for predicting retail sales, the performance of dummy regression models may not be robust. Furthermore, trigonometric models are not useful in aggregate retail sales forecasting.
Article
The difficulty with fashion retail forecasting is due to a number of factors such as the season, region and fashion effect and causes a nonlinear change in the original sales rules. To improve the accuracy of fashion retail forecasting, a two-stage dynamic forecasting model is proposed, which is combined with both long-term and short-term predictions. The model introduces the improved adjustment methods, the main adjustment model and error forecasting model in the adjustment system collaborated with each other. The real-time data are demonstrated by applying the model in wireless mobile environment. The experiment shows that the model provides good results for fashion retail forecasting.
Article
The objective of this paper is to inform the information systems (IS) manager and business analyst about the role of machine learning techniques in business data mining. Data mining is a fast growing application area in business. Machine learning techniques are used for data analysis and pattern discovery and thus can play a key role in the development of data mining applications. Understanding the strengths and weaknesses of these techniques in the context of business is useful in selecting an appropriate method for a specific application. The paper, therefore, provides an overview of machine learning techniques and discusses their strengths and weaknesses in the context of mining business data. A survey of data mining applications in business is provided to investigate the use of learning techniques. Rule induction (RI) was found to be most popular, followed by neural networks (NNs) and case-based reasoning (CBR). Most applications were found in financial areas, where prediction of the future was a dominant task category.
Article
Although today's world offers us unprecedented access to greater and greater amounts of electronic information, we are faced with significant problems when it comes to finding the right information at the right time-the essence of the information-overload problem. One of the proposed solutions to this problem is to develop technologies for automatically learning about the implicit and explicit preferences of individual users to customize and personalize the search for relevant information. In this article, we describe the development of the personalized television listings system (PTV),1 which tackles the information-overload problem associated with modern TV listings data by providing an Internet-based personalized TV listings service so that each registered user receives a daily TV guide that has been specially compiled to suit his/her particular viewing preferences.