Highlights in Science, Engineering and Technology AMMSAC 2024

Volume 107 (2024)

Walmart Sales Prediction Using Multiple Linear Regression

Anting Lin*
Department of economic, Hangzhou Dianzi University, Hangzhou, 310000, China
*Corresponding author: 22150136@hdu.edu.cn
Abstract. The sales of Walmart are often considered to be related to many factors, and indeed this
is the case. This study aims to examine the impact of objectively observable independent variables
on Walmart's weekly sales. The paper employs the Z-score standardization technique to normalize
the dimensions and utilizes linear regression analysis to process data from 6,435 weekly sales
records of Walmart. The model is derived after preprocessing and standardizing the data in this
paper. The findings reveal that most variables have either a minor or significant effect on sales.
Notably, temperature and Fuel Price exert a certain influence on sales, while Holiday, Consumer
Price Index (CPI), and Unemployment rates have a substantial impact. This investigation into the
objective factors affecting Walmart's sales will benefit the management by providing insights to refine
strategic planning through environmental surveys, thereby aiding in the enhancement of the
shopping atmosphere and customer experience, ultimately aiming to boost sales revenues.
Keywords: Walmart weekly sales; linear regression; objective influencing factors.

1. Introduction
Shopping is an indispensable part of modern people’s daily life and sales are regarded as one of
the important indicators of supermarket work efficiency. As one of the largest supermarkets in the
world, Walmart's sales often receive widespread attention. As of January 2023, Walmart is the largest
retailer in the world, with 10,623 stores and 380 distribution facilities in 27 countries and its revenues
since 2022 have exceeded $600 billion a year [1]. In recent two years, Walmart U.S. comp sales
increased 19.6% and eCommerce sales grew 35% [2]. Obviously, there are a lot of objective factors
that affect Walmart's sales. Exploring the impact of various factors and predicting sales will also
become the most important decision-making reasons for Walmart managers. Therefore, this paper
aims to find out objective factors that have a greater impact on sales and build a mathematic model
for forecasting Walmart sales.
The factors that affect the sales of Walmart are many and complex, involving a variety of
subjective and objective factors. Niu put forward the XGBoost sale prediction model which combines
XGBoost algorithm and meticulous feature engineering processing for predicting Walmart's sales
problem [3]. The experimental results indicate that the sale forecast model based on XGBoost
outperforms the other machine learning models. Xiao used multiple linear regression methods and
built deep neural network models to analyze data and make predictions [4]. In this paper, in order to
avoid reducing the multicollinearity problem, the correlation between attributes is taken into account,
explaining the stronger correlation features between variables. At the same time, the concept of
dimension is also introduced for data preprocessing. Singh et al. adopted big data analysis, including
Spark and MapReduce [5]. In particular, those researchers leverage the Spark framework's Scala and
Python API to understand Walmart's sales action strategy and explain consumer behavior through
data visualization. Finally, under the analysis of factors such as temperature, fuel prices and holidays,
and it is pointed out that more sales can be generated when fuel prices are within a reasonable range,
for example.
In similar directions, Yao found the prediction effect of Random Forest Registrar works well and
the study results shed light on guiding further exploration of Sales forecasts for supermarkets [6]. Wu
et al. illustrated that Auto-Regressive Moving Average Model (ARMAV) with linear trend method
is the best benchmark model to forecast sales data for new product with trend and with sales person's
inputs [7]. The EMD-XGBOOST-RLSE model is used to solve the complex production process
prediction problem by Xu et al. [8]. Harsoor et al. look at Walmart's sales at stores in different
124
Highlights in Science, Engineering and Technology AMMSAC 2024
Volume 107 (2024)

geographic locations [9]. Some people think that the improved least squares support vector machine
algorithm can effectively predict business information [10]. Jeswani’s report shows that the gradient
lift model provided the most accurate sales predictions and slight relationships were observed
between factors such as store size, holidays, unemployment, and weekly sales [11].
In summary, after consideration and optimization, this paper will analyze five variables:
Temperature, Holiday Flag, Fuel Price, customer price index (CPI), Unemployment to build a linear
regression model for predicting Walmart sales.

2. Methods
2.1. Data Source
The dataset used in this paper is fetched from the Kaggle website (Walmart Dataset). It was from
2010 to 2012, in the file Walmart Store sales. This dataset contains 6435 groups of data, and this
research selected all of them as samples. The original dataset remained in .csv format.
2.2. Variable Selection
The original dataset has a large amount of data. But there is no null for variables, so this literature
directly preprocessed the dataset. By pre-processing the above variables through Z-score
normalization, a dataset with a unified dimension can be obtained. The formula for Z-score
normalization is provided here:
𝑥𝑖 −𝑥̅
𝑧= (1)
𝑆𝑡𝑑

The dataset contains 5 variables (Temperature, Holiday Flag, Fuel Price, customer price index
(CPI), Unemployment) and one dependent variable (Weekly sales). The specific description of this
dataset is shown in Table 1:
Table 1. List of Variables
Variable Logogram Meaning
Temperature 𝑥1 Temperature on the day of sale
Holiday Flag 𝑥2 Some special period may deeply influence sales
Whether the week is a special holiday week 1 – Holiday week
Fuel Price 𝑥3
0 – None holiday week
CPI 𝑥4 Prevailing consumer price index
Unemployment 𝑥5 Prevailing unemployment rate
Weekly sales Y Sales for the given store

2.3. Method Introduction

This paper employs the method of multiple linear regression for analysis. The initial effort focused
on identifying relationships among various variables, which were then used to construct a linear
regression model for predictive purposes.
Multiple linear regression is a statistical method used to analyze the linear relationship between
two or more independent variables (explanatory variables) and one dependent variable (response
variable). The multiple linear regression model describes the relationship between the independent
and dependent variables through an equation that typically includes an intercept term and multiple
slope parameters, each corresponding to an independent variable. The model equation is as follows:
𝑦 = 𝛽0 + 𝛽1 𝑥1 + 𝛽2 𝑥2 + ⋯ + 𝛽𝑝 𝑥𝑝 + ε (2)
And the parameters in the model, including the intercept and slopes, are usually estimated using
the least squares method. This approach aims to minimize the difference between the observed values
125
Highlights in Science, Engineering and Technology AMMSAC 2024
Volume 107 (2024)

and the predicted values of the model. After estimating the model parameters, significance tests are
commonly performed to determine if the independent variables have a statistically significant effect
on the dependent variable. To ensure the validity of the model, regression diagnostics are conducted
to check whether the data adheres to the basic assumptions of the model, such as the independence
of error terms, normality, and homogeneity of variance. Overall, multiple linear regression is a
powerful tool that helps researchers understand complex relationships among multiple variables and
plays a role in prediction and decision-making.

3. Results and Discussion

3.1. Descritive Analysis
Initially, a box plot of Walmart's weekly sales can be generated, which provides a rough idea of
the range of its data. Below is the resulting box plot presentation:

Fig. 1 Walmart Weekly Sales box plot

As can be seen from the Fig. 1, Walmart's weekly sales are mostly distributed around 1,000,000,
roughly exhibiting a skewed normal distribution. It can also be observed that there are many outliers
in the data, indicating a need for preprocessing of the data.

Fig. 2 Weekly Sales & Holiday Flag box plot

126
Highlights in Science, Engineering and Technology AMMSAC 2024
Volume 107 (2024)

In Fig. 2, it's evident that Walmart's sales during holidays are significantly higher than on non-
holidays, which aligns with common sense. In summary, the factors affecting Walmart's weekly sales
are diverse and complex, encompassing both subjective and objective elements. This paper focuses
on studying the impact of objective factors on Walmart's weekly sales. However, to avoid the issues
of interaction between objective factors and multicollinearity, a correlation analysis will be conducted
among the independent variables below.
3.2. Model Results
Fig. 3 is a heatmap of Pearson correlation coefficients. Clearly, there are no very strong
correlations among the independent variables studied in this paper. The Pearson correlation
coefficients between each pair of random variables do not exceed 0.3, indicating that there is no strong
linear relationship between the variables, thus avoiding the problem of multicollinearity.

Fig. 3 Pearson correlation coefficient

Table 2. Regression coefficient table

Β S.E. Beta t p VIF Tolerance

Constant -0.009 0.013 - -0.726 0.468 - -
Temperature -0.024 0.013 -0.024 -1.808 0.071 1.130 0.885
Holiday Flag 0.133 0.049 0.034 2.710 0.007**1.029 0.972
Fuel Price -0.008 0.013 -0.008 -0.645 0.519 1.084 0.922
CPI -0.111 0.014 -0.111 -8.194 0.000**1.221 0.819
Unemployment -0.138 0.013 -0.138 -10.4600.000**1.150 0.869
R2 0.025
F F (5,6429) =33.570, p=0.000
D-W 0.114

Table 2 shows the regression coefficients for the multiple linear regression equation model. It is
evident that the p-values for Holiday Flag, CPI, and Unemployment are less than 0.05, indicating that

127
Highlights in Science, Engineering and Technology AMMSAC 2024
Volume 107 (2024)

these factors have a significant impact on Walmart's weekly sales. Moreover, based on the above data,
the relevant multiple linear regression equation can be derived and compared:
𝑦 = −0.009 − 0.024𝑥1 + 0.133𝑥2 − 0.008𝑥3 −0.111𝑥4 − 0.138𝑥5 + ε (3)
This model was derived after preprocessing and standardizing the data in this paper, with a focus
on studying the correlation between various independent variables and dependent variables. At the
same time, it has also minimized the issue of multicollinearity to a considerable extent, making it
more aligned with real scenarios.

4. Conclusion
The research presented in this paper focuses on examining the correlations between various
variables, aiming to describe these relationships through the methodologies of linear regression. The
findings indicate that Walmart's weekly sales are influenced by factors such as temperature, holidays,
fuel prices, consumer price index (CPI), and unemployment rate. Among these, holidays, CPI, and
the unemployment rate have a more substantial impact on sales.
It is undeniable that the factors affecting Walmart's weekly sales are diverse and complex. This is
not only due to objective factors but also many subjective elements that alter sales to some extent,
such as the service attitude of store employees and local shopping preferences. Due to the limitations
in data volume and a lack of subjective awareness regarding the impact factors, it is challenging for
linear models to provide a perfect explanation for the prediction of actual results. This also
significantly affects the accuracy of the model. Nevertheless, this paper still possesses certain
strengths and feasibility. On one hand, the adopted method of linear regression allows for a more
direct presentation of how each independent variable impacts the dependent variable, with correlation
analysis effectively demonstrating the trends in the dependent variable and predicting its trajectory.
On the other hand, the use of visual graphics to illustrate the data makes the results more intuitive.
The purpose of this paper is to provide Walmart operators with suggestions regarding the factors
that influence Walmart's sales. The objective factors studied in this paper will assist them in
understanding the current operating environment and timely improve their business strategies,
thereby enhancing selling efficiency and offering customers a better shopping experience.

References
[1] Hu Xiao. Wal-Mart Retail Sales Data Forecast Analysis. Statistics and Appliciations, 2023, 12(6): 1683-
1695.
[2] Singh M, Ghutla B, Jnr R L, et al. Walmart's Sales Data Analysis - A Big Data Analytics Perspective.
2017 4th Asia-Pacific World Congress on Computer Science and Engineering (APWC on CSE), 2017.
[3] Yao B. Walmart Sales Prediction Based on Decision Tree, Random Forest, and K Neighbors Regressor.
Highlights in Business, Economics and Management, 2023, 5: 330-335.
[4] Wu L, Yan J Y, Fan Y J. Data Mining Algorithms and Statistical Analysis for Sales Data Forecast. Fifth
International Joint Conference on Computational Sciences & Optimization, IEEE, 2012, 577-581.
[5] Xu Z G. Complex production process prediction model based on EMD-XGBOOST-RLSE. Proceedings
of 2017 9th International Conference on Modelling Identification and Control (ICMIC), 2017.
[6] Harsoor A S, Patil A. Forecast of Sales of Walmart Store Using Big Data Applications. Working paper,
2024.
[7] Zhou M, Wang Q. The On-line Electronic Commerce Forecast Based on Least Square Support Vector
Machine. International Conference on Information & Computing Science. IEEE, 2009.
[8] Jeswani R. Predicting Walmart Sales, Exploratory Data Analysis, and Walmart Sales Dashboard.
Rochester Institute of Technology, 2021.
[9] Zhu Xiaohua. Impact and Analysis of the Implementation of the New Income Standards on the
Recognition of Revenue in Cooperative Supermarkets. Shanghai Business, 2023, 2: 34-36.

128
Highlights in Science, Engineering and Technology AMMSAC 2024
Volume 107 (2024)

[10] Left Liangjun, Zhang Lijuan. Analysis of the Impact of Agricultural Supermarket Operation on the
Agricultural Industry Chain. Rural Economy, 2003, 2.
[11] Bin Murong, Zhou Faming. Economic analysis and promotion ideas for the operation of agricultural
product chain supermarkets. National Business Situation and Economic Theory Research, 2007.

129