A Structured Literature Review on the Application of Machine Learning

in Retail

Marek H

utsch and Tobias Wulfert

Chair of Business Informatics and Integrated Information Systems, University of Duisburg-Essen, Essen, Germany

Keywords:

Retail, Electronic Commerce, Machine Learning, Application Area, Literature Review.

Abstract:

Machine learning (ML) has the potential to take on a variety of routine and non-routine tasks in brick-and-

mortar and e-commerce. Many tasks previously executed manually are susceptible to computerization in-

volving ML. Although procedure models for the introduction of ML across industries exist, it needs to be

determined for which tasks in retail ML can be implemented. Hence, we conducted a structured literature

review involving 225 research papers to derive possible ML application areas in retail along with the structure

of a well-established information systems architecture. In total, we identiﬁed 20 application areas for ML in

retail that mainly address decision-oriented and economic-operative tasks. We organized the application areas

in a framework for practitioners and researchers to determine an appropriate ML usage in retail. Our analysis

further revealed that while ML applications in ofﬂine retail focus on the article, in e-commerce the customer

is pivotal for application areas of ML.

1 INTRODUCTION

It is estimated that by 2025 more than 150,000 mo-

bile robots will be deployed in retail to execute rou-

tine work such as reﬁlling shelves or preparing pack-

age dispatch making use of machine learning (ML)

(Liang et al., 2019).However, ML cannot only be im-

plemented to overtake cognitive routine tasks involv-

ing rule-based tasks (Goos et al., 2009), but can also

spread into domains previously deﬁned as non-routine

tasks (e.g., handwriting recognition or car driving)

(Veres et al., 2011). A recent study by Frey and Os-

borne (Frey and Osborne, 2017) identiﬁed that many

tasks in online and ofﬂine retail are susceptible to

computerization. Although retail in both ofﬂine and

online environments involves a variety of tasks requir-

ing different capabilities, a recent study by McKin-

sey revealed a high potential for automatizing in gen-

eral and ML in retail (Manyika et al., 2017). The

focus in retail is on primary and valuable tasks that

are described by Sch

utte as economic-operative tasks

(Sch

utte, 2017; Manyika et al., 2017)

Brick-and-mortar retailers have faced increasing

competitive pressure in recent years, especially from

online retailers. This process has been intensiﬁed in

particular by the COVID-19 pandemic (Nicola et al.,

https://orcid.org/0000-0002-5504-0718

2020). While the use of ML has become strongly es-

tablished in e-commerce, the pressure on brick-and-

mortar retailers is also increasing in this area (Große

Holtforth, 2018). While brick-and-mortar retailers

manage products, logistics, etc. on store level, online

retailers focus on optimizations on the customer level

(Sch

utte, 2017). E-commerce involves much data re-

garding its customers and unstructured data regard-

ing their products (product reviews, products created

by users on a marketplace) the situation in brick-and-

mortar retail is different (D’Haen et al., 2012; H

utsch

and Wulfert, 2021; Niu et al., 2017). The majority of

data is present in a structured manner and data about

articles is often self-managed by product owners. Ad-

ditionally, there is only a limited amount of data about

the customers potentially limiting the introduction of

ML to only a few application areas.

The ML algorithm selection problem - also known

as the combined algorithm selection and hyperparam-

eter optimization (CASH) problem - poses a chal-

lenge to all ML practitioners (Thornton et al., 2013).

This not only involves the selection of an algorithm

that ﬁts the data set and the economic problem of

the retailer but also model selection and hyperparam-

eter optimization (Biem, 2003). Due to the abun-

dance of possible combinations of these factors and

their inﬂuence on the quality of the result, it is essen-

tial to choose ﬁtting factors (i.e. algorithm, model,

332

Hütsch, M. and Wulfert, T.

A Structured Literature Review on the Application of Machine Learning in Retail.

DOI: 10.5220/0011043200003179

In Proceedings of the 24th International Conference on Enterprise Information Systems (ICEIS 2022) - Volume 1, pages 332-343

ISBN: 978-989-758-569-2; ISSN: 2184-4992

and hyperparameter) that ﬁt the underlying data as

well as possible (Thornton et al., 2013). Recently,

however, there have been created various automated

ML tools for making these decisions, which greatly

facilitate the real-world application of ML methods

(Feurer et al., 2019). Thus, even practitioners who are

non-experts in the various ML disciplines can achieve

good results using automated ML tools and avail-

able procedure models for introducing ML. Although

these tools exist, it is still necessary to have experts

who determine which tasks can be supported or even

automated by ML and who are able to create the in-

creasingly complex ML models themselves (H

utsch

and Wulfert, 2021). ML experts can be supported by

industry-speciﬁc frameworks indicating possible ML

applications. Such a framework can resemble an en-

try point for ML implementation in online and ofﬂine

retail. However, existing literature reviews on the ap-

plication of ML in retail are usually limited to one

application area and focus only online or ofﬂine re-

tail (Bousqaoui et al., 2018). Against this backdrop,

we pose the following research question: What tasks

in (online and ofﬂine) retail can be supported by ma-

chine learning? To address this research question, we

provide an overview of multiple application areas for

ML in retail. We conducted an application-centered

literature review for extracting possible ML usages

in retail (Vom Brocke et al., 2009). We make use of

an agreed-upon reference model in retail to structure

the identiﬁed ML application areas along with retail-

speciﬁc tasks (Sch

utte, 2017). This paper is useful for

the development of a customized strategy of ML in re-

tail. Hence, we propose decision support for retailers

the ML usage in speciﬁc application areas.

The remainder of this research paper is structured

as follows: in section 2 we elaborate on fundamentals

about retail information systems in general and ML

in particular. Next, we elicit our structured literature

review approach. While we present the identiﬁed ap-

plication areas of ML in retail in section 4, we discuss

ML’s current and future applications in online and of-

ﬂine retail in section 5. This research closes with a

brief conclusion and research outlook.

2 THEORETICAL BACKGROUND

Next, we discuss related litureture to retail informa-

tion systems and machine learning.

2.1 Retail Information Systems

Since the ignition of the internet existing retailers

either extended their traditional in-store business by

an additional online channel or new companies were

started up implementing an e-commerce business

model and selling articles online without operating

any brick-and-mortar stores (Rudolph et al., 2015).

Regardless of the institutional-economic discussion

of retail companies, the tasks intended with the trad-

ing functions are economically necessary (Laudon

and Traver, 2020). These are the three basic functions

bridging the discrepancy between manufacturer and

customers in the streams in real goods (goods, ser-

vices; returns), nominal goods (money, credits), and

information across space, time, quantity, and quality

(Barth et al., 2015). Facilitated by the ongoing digi-

talization of the retail sector, the three basic functions

increasingly cope with digital product and price in-

formation, payment, logistics, and distribution pro-

cesses (Levy et al., 2019). In e-commerce, trading

transactions are carried out digitally to some degree

(Laudon and Traver, 2020). The degree to which the

transaction must be digital varies in the literature on

a continuum between a completely digital transac-

tion and only a small part of the procurement pro-

cess (Laudon and Traver, 2020). Information sys-

tems in retail support the execution of the three trad-

ing functions and related tasks. They support the

operational-dispositive, the business administration-

administrative, and the controlling as well as corpo-

rate planning tasks (Becker and Sch

utte, 2004). They

extend the components of merchandise management

systems (merchandise planning, logistics, and settle-

ment processes) by business intelligence and neces-

sary corporate-administrative tasks in an integrative

manner for carrying out the business processes of a

retail company (Becker and Sch

utte, 2004; Sch

utte,

2017). The necessary tasks of a retailer and support-

ing information systems can be depicted in process

models and reference models for the retail industry.

Reference models comprise a high-level sketch of the

system and application architecture of a speciﬁc com-

pany and part of its application architecture (Heinrich

and Stelzer, 2009). In the current body of literature

exists a number of reference models describing re-

tail information systems (Becker and Sch

utte, 2004;

Sch

utte, 2011; Aulkemeier et al., 2016b). In this ar-

ticle, we focus on the shell model as proposed by

Sch

utte as it covers the whole value chain from manu-

facturer, wholesaler, and retailer to the end consumer,

focuses on tasks on a business level, and is designed

for both online and ofﬂine environments. The shell

model is a task-oriented reference model for retail

(Sch

utte, 2011). We use the task-oriented perspective

to determine for which tasks ML is already applied

in scientiﬁc literature. It follows the principle of sep-

aration of tasks and actors introduced by Ferstl and

A Structured Literature Review on the Application of Machine Learning in Retail

333

Sinz (Ferstl and Sinz, 2013). The shell model copes

with the identiﬁed shortcomings of the H-model such

as the artiﬁcial separation between merchandise man-

agement and decision support systems. It consists of

four separate but intertwined architectures for each of

the aforementioned actors along a value chain. The

retail architecture consists of ﬁve shells for the main

retail tasks (master data, technical tasks, economic-

operative tasks, administrative tasks, and decision-

oriented tasks) of each actor and the retailer in par-

ticular (Sch

utte, 2017). Each shell consists of a series

of tasks that form the components of the architecture.

The tasks relevant for structuring the application ar-

eas are described in the analysis section (Section 4).

The shell model is depicted in Figure 1.

Figure 1: Retail information system architecture (Sch

utte,

2011; Sch

utte, 2017).

2.2 Machine Learning

Historically, computerization has largely been con-

ﬁned to manual and cognitive routine tasks involv-

ing explicit rule-based activities (Goos et al., 2009).

Following recent technological advances, ML is now

spreading to domains commonly deﬁned as non-

routine (Veres et al., 2011). As these non-routine

tasks are commonly executed by human employ-

ees, this paper reviews research articles to investigate

which of these non-routine tasks in ofﬂine and online

retailing are support by ML in research and thus are

potential candidates to support humans with ML in

the future. We use the following deﬁnition of ML in

which “a computer program is said to learn from ex-

perience E with respect to some class of tasks T and

performance measure P, if its performance at tasks

in T, as measured by P, improves with experience E”

(Mitchell et al., 1997). To fulﬁll this deﬁnition an ML

algorithm needs an optimization algorithm, an error

function, a model and a data set. As we use some ML

vocabulary in the remaining paper we clarify them

in the following. A forecast is deﬁned by historical

time-series data that is used to make assumptions of

possible events in the future. A classiﬁcation algo-

rithm makes statements about a data set and assigns

this data to a given set of classes/ labels. There are a

lot of other ML tasks which are mostly incorporated

with a combination of forecasts and classiﬁcation al-

gorithms, which we mostly referred to as ”analysis”

in the following. In a similar work, Weber and Sch

utte

(Weber and Sch

utte, 2019) inductively determined ar-

tiﬁcial intelligence use cases for brick-and-mortar re-

tail from real-world examples. Although the broader

term of artiﬁcial intelligence is used there, application

areas of ML are also included. Just as in this paper,

Weber and Sch

utte (Weber and Sch

utte, 2019) use the

shell model as a starting point for the literature re-

view (Figure 1). Both papers complement each other

by identifying ML applications from practical contri-

butions (Weber and Sch

utte, 2019) and from analyz-

ing scientiﬁc contributions. The different approach

of this paper makes sense, as the underlying research

question is another one. In this paper potential appli-

cation areas of ML in retail are reviewed by conduct-

ing research papers as a framework for practitioners

and overview of researchers. Another related work

by Bousqaoui et al. (Bousqaoui et al., 2018) focused

on the use of ML in the domain of supply chain man-

agement.

3 SCIENTIFIC APPROACH

For identifying relevant application areas of ML in re-

tail (online and ofﬂine), we followed the structured lit-

erature review approach as proposed by vom Brocke

et al. (Vom Brocke et al., 2009). Following Cooper

(Cooper, 1988) our literature review can be charac-

terized as follows: We focus on research outcomes

(use cases and ML algorithms for retail) and applica-

tions of ML in retail. We aim to integrate the body of

literature and generalize application areas from avail-

able research. We neutrally represent articles from

our exhaustive literature review with citations of se-

lected papers for each application area. In this paper,

we collect results from previous works and integrate

them into an overarching framework. Our research

addresses scholars specialized in the use of ML in re-

tail and e-commerce and practitioners implementing

ML algorithms.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

334

We identify keywords from related literature and

an initial literature screening resulting in a generic

search query that consists of a retail and an ML part

(Figure 2). As we derive existing application areas of

ML in retail from relevant journals and conference pa-

pers and propose avenues for further ML application

in this domain, we choose a broad literature scope.

This broad scope is underpinned by the search en-

gines SCOPUS, Web of Science (WoS), IEEE Xplore

(IEEE), AIS electronic library (AISeL), and ACM,

we query with their underlying databases that con-

sist of research from, among others, economics and

computer science literature. An initial query for ti-

tle and keywords on 2021-04-15 resulted in an ini-

tial sample of 5,538 papers. To ensure an appropriate

level of quality we focus on scientiﬁc literature and

added additional quality criteria to the search (Ran-

dolph, 2009). We excluded non-English and non-

German language articles; panels, and commentaries;

purely technical articles (e.g., articles that focus ex-

clusively on technological aspects without applying

them in an e-commerce context); and articles with a

pure e-commerce focus (e.g., articles that focus ex-

clusively on e-commerce or sub-types without the in-

clusion of conceptual models). In contrast to previ-

ous research, such as Weber and Sch

utte (Weber and

Sch

utte, 2019), we intentionally excluded white pa-

pers and practitioner reports. Based on the title and

abstract considering the quality criteria, we reduce

our sample to 410 papers. Further applying the ap-

proach of Bandara et al. (Bandara et al., 2015) to

the content of the papers, we identiﬁed 261 relevant

publications, leaving 197 after the exclusion of dupli-

cates. We use a one-time for- and backward search

to identify research that is presented as an alternative

scenario resulting in 28 additional papers forming a

ﬁnal set of 225 papers (Figure 2).

Figure 2: Structured Literature Review Design.

The articles within the ﬁnal set were indepen-

dently analyzed by full-text screening and the rele-

vant text passages including possible application ar-

eas were coded to reﬁne the set of application ar-

eas. The shell model with its central master data

objects (ﬁrst shell), technical tasks (second shell),

operational tasks (third shell), administrative tasks

(fourth shell), and decision-oriented tasks (ﬁfth shell)

(Sch

utte, 2011) served as the basis for the coding,

which was inductively adapted as needed. The ini-

tial set of possible application areas was reﬁned using

our ﬁnal set of papers by two independent researchers

in three coding rounds. After each round, the re-

searchers met via online communication media to dis-

cuss the application area reﬁnements. The framework

in Figure 3 depicts the 20 application areas for ML in

retail.

4 APPLICATION AREAS OF

MACHINE LEARNING

In the following the application areas for ML are pre-

sented, which result from our structured literature re-

view. The assignments between business tasks of the

retail shell model and ML application area are visu-

alized in ﬁgure 3. The number of identiﬁed scientiﬁc

papers is visualized in brackets in Figure 3 for each

application area. Each application area is assigned to

a task of the shell model. We present each application

area in general and the results of the literature review

are described speciﬁcally by outlining a sample of the

literature review results.

4.1 Decision-oriented Tasks

The decision-oriented tasks (i.e., business intelli-

gence) of the shell model provide aggregated infor-

mation to the management of a retailer (Sch

utte and

Vetter, 2017). One very important business intelli-

gence ML application area is the sales forecast, which

is represented by 72 papers. The sales forecast pro-

vides management with information on future cus-

tomer demands and therefore, management is able to

plan operating activities accordingly.

Sales forecasting per article is one of the ML ar-

eas in retail that receives the most attention, because

it is the basis for numerous advanced algorithms,

such as price optimization (Chandrashekhara et al.,

2019) or promotion optimization (Henzel and Sikora,

2020). Forecasting techniques are divided into four

main groups: qualitative methods which are based on

human judgment, time-series methods that use his-

torical data, causal methods use rule-based forecasts,

and simulation methods that simulate the behavior of

the customer (Chopra et al., 2013). If the focus is

on-demand forecasting with ML methods, time series

methods are mostly used. The forecast for individ-

ual days has the difﬁculty that rare events are deci-

sive in the prognosis, which are often hardly consid-

ered as outliers on weeks or monthly levels (Huber

and Stuckenschmidt, 2020). The demand forecast can

be extended by features, for example, Verstraete et al.

A Structured Literature Review on the Application of Machine Learning in Retail

335

Business Intelligence

Handing-out Goods

Managing Goods

Serving Customers

Demand and

Sales Forecast

(72)

Review

Consistency (2)

Reclamation Forecast (1)

Fake Review

Detection (5)

Customer

Churn(4)

Customer

Repurchase (9)

Customer

Conversion (19)

Review

Classification

(16)

Decision

-oriented

Tasks (4.1)

Economic

-operative Tasks (

4.2)

Robotic Shopping

Assistant (2)

Retailer/Supplier

Analysis (6)

Article Prices (6)

Promotions (4)

Site Selection (10)

Unmanned Checkout

System (12)

Fraud

Detection

(16)

Function (3)

Product

Classification (12)

Product

Recommendation

(17)

Lingual Shopping

Assistant (4)

Ordering Goods

Out of Stock Detection

(6)

Figure 3: Framework for the Machine Learning Application in Retail.

(Verstraete et al., 2019) create a short-term and long-

term forecast model that takes weather data into ac-

count. Loureiro et al.(Loureiro et al., 2018) compare

demand forecasts in the context of the fashion indus-

try, where short product life cycles take place.

The inverse of a sales forecast is the forecast of

returned goods. For example, when it comes to prod-

uct recalls, a high volume of returned products must

be factored into operational activity planning by man-

agement. In addition, the returns process deals with

goods that have been complained about. This can

happen in case of wrong deliveries, faulty goods, or

even defective goods (Becker and Sch

utte, 2004). Ku-

mar et al. (Kumar et al., 2014) used multiple ML al-

gorithms for the reclamation forecast. This forecast

is integrated into supply chain design and planning

problems to optimize supply chain tasks.

A store is an organizational unit that is used to

document the goods in stock for inventory manage-

ment and to map the associated business processes,

such as goods receipt, physical inventory, and goods

issue (Becker and Sch

utte, 2004). Part of the pro-

cess for opening a new store includes the choice of

location (site selection), which is done by manage-

ment, where ML algorithms can support. For exam-

ple, Wang et al.(Wang et al., 2018) make a quanti-

tative site selection, so that a very large number of

possible store locations are reduced to a small num-

ber, which is then investigated further by qualiﬁed

employees. Machado et al.(Machado et al., 2015)

follow a different approach in which a pre-selection

has already been made and the remaining ﬁve stores

are evaluated by a neural network. Another formu-

lation of the site selection problem is when a mall

is designed with a given set of sets. For example,

Miao(Miao, 2020) uses ML algorithms to ﬁnd a rea-

sonable layout for shopping malls where the relation-

ships of stores to each other are taken into account.

4.2 Economic-operative Tasks

The next subsections address the economic-operative

tasks of the shell model, which are the primary ac-

tivities of retailers. These tasks fulﬁll the bridging

functions and resemble a retailer’s value proposition

(Levy et al., 2019). We have identiﬁed ML applica-

tion areas for the tasks of managing goods, serving

customers, hand out goods, and goods ordering.

4.2.1 Managing Goods

A central aspect of the economic-operative tasks is

the managing of goods, which is reﬂected by a large

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

336

number of papers (59) in this area. The task managing

goods consists of the assortment policy, conditions

policy (purchasing price and selling price), placement

policy, and promotion policy. Prices are an important

driver to inﬂuence how often an article is sold and the

resulting proﬁt margin. Therefore retailers can imple-

ment price optimization so that the demand ﬁts the

supply. For example, Chandrashekhara et al. (Chan-

drashekhara et al., 2019) determine the best prices

of smartphones considering different features of the

smartphones. Furthermore, promotion optimization

can be used to speciﬁcally increase sales for a short

period of time. The general promotional effect is

forecasted by Henzel et al.(Henzel and Sikora, 2020)

using. Retailers implement product recommendation

systems to enhance the matching between supply- and

demand-side participants, which we assign to promo-

tional tasks and thus the task of the managing good.

These systems recommend supply-side products to

the customers based on previous purchases, ratings,

or search behavior (Katarya and Verma, 2017). In

contrast to the traditional recommendation that sug-

gests popular products to customers, their algorithm

aims at matching customers with appropriate niche

products. Thus, online shops can more speciﬁcally

target the long tail (McAfee and Brynjolfsson, 2017)

and exploit niche markets to generate additional re-

vue. For determining the ranking of products that

should be recommended to demand-side participants,

Liang et al. (Liang and Wang, 2019) perform sen-

timent analyses on analyzing online product reviews

from Taobao.com. As personalized product recom-

mendations may be biased by supply-side marketing

endeavors, Wan et al. (Wan et al., 2020) propose ML

algorithms to allow product recommendation in un-

derrepresented market segments. The authors used

two data sets from online shops selling apparel and

electronics. Promotional efforts can be done for a

special group of customers. To detect such a group

of customers ML algorithms are used, which we also

categorized to the promotional task.

An important application area of ML in retailing

is customer churner classiﬁcation (Li and Li, 2019).

Clustering churners categorize demand-side partici-

pants into two clusters: one with churning customers

and the other one with non-churning customers which

results in a binary classiﬁcation problem (Kim and

Lee, 2012). While possible churners can be addressed

with special marketing campaigns, the quality of ser-

vices can be targeted at non-churning customers. Cus-

tomer churn describes the loss of customers (Kim and

Lee, 2012), which generated sales in the past and will

no longer generate sales, mostly through an active de-

cision to generate sales with a competitor. Keeping

existing customers in the sales funnel is considered

less expensive than acquiring new customers (Singh

and Agrawal, 2019; Nikulin, 2016). Customer sat-

isfaction has a positive inﬂuence on customer repur-

chase behavior(Jheng and Luo, 2019). To implement

a customer repurchase analysis Kumar et al.(Kumar

et al., 2019) used a combination of ML techniques

by using customer characteristics and e-commerce at-

tributes. This application area allows retailers to pre-

dict if a demand-side participant will rejoin the sales

funnel resulting in additional sales transactions and

revenue.

Using an ML algorithm on e-commerce data

Singh et al.(Singh and Agrawal, 2019) suggest de-

tecting very loyal customers and provide them with

higher quality services to increase their satisfaction to

retain them in the sales funnel. Jheng et al. (Jheng

and Luo, 2019) mine transaction logs of an enterprise

resource planning system at an online shop to pre-

dict customer repurchase behavior. Moreover, Ma-

haboob et al.(Mahaboob Basha et al., 2020) develop

an ML model to describe the moderating role of cus-

tomer loyalty on customer retention. Predicting the

probability of a lead to convert into a customer and

the proﬁtability of this new customer is crucial in re-

tailing (D’Haen et al., 2012). The customer conver-

sion analysis implemented by D’Haen et al.(D’Haen

et al., 2012) applies ML algorithm on sales data of

a German B2B e-commerce company and comple-

mentary data crawled on the internet to predict the

proﬁtability of a lead. Knowing which leads will be

proﬁtable can support retailers by directly address-

ing these leads and increase their conversion proba-

bility. Niu et al.(Niu et al., 2017) apply multiple ML

algorithms to predict the probability of a customer

purchasing a product based on his (prior) search be-

havior. The computational model was applied to e-

commerce data from Walmart. Thus, a retailer can

predict the customer’s willingness to purchase and

pay in an early stage of a sales funnel (Blank and

Dorf, 2012).

4.2.2 Serving Customers

The following application areas are concerned with

tasks that serve customers. In brick-and-mortar

stores, the products are made available through lim-

ited shelf space. The digital equivalent is the online

stores, which often provide a much larger assortment.

Therefore, a product search function is necessary to

serve customers with their demand, which is the rea-

son which we categorized the product search function

to the business task. Khatwani et al.(Khatwani and

Srivastava, 2016) develop a model for predicting cus-

tomer’s individual information search preferences us-

A Structured Literature Review on the Application of Machine Learning in Retail

337

ing questionnaire data.

To enable a good product search, products must

be classiﬁed ﬁrst. E-commerce product classiﬁca-

tion is challenging due to the large scale and com-

plexity of the product information and categories. Yu

et al.(Yu et al., 2018) combine multiple ML algo-

rithms to propose e-Commerce Text Classiﬁcation.

As reviews are the online counterpart to recommen-

dations of an employee, we categorize review anal-

ysis with ML algorithms to the economic-operative

tasks serving customers. To propel a retailer’s role

as a trustee they often implement systems to review

the transaction partner and the products sold. Suppli-

ers and customers of a retailer are aware of reviews

as they often directly impact their business (Hussain

et al., 2020). Vinodhini and Chandrasekaran (Vinod-

hini and Chandrasekaran, 2014) implemented based

on the customers opinion (opinion mining) for review

classiﬁcation. They used publicly available customer

reviews to classify the reviews into positive and nega-

tive reviews. So supply- and demand-side participants

can preselect reviews possibly more relevant for them.

Opinion spammers and fake reviews exploit cus-

tomer trust and harm the reputation of the retailer

in e-commerce as a trustee by posting false or de-

ceptive reviews (Zhang et al., 2016). These re-

views are difﬁcult to detect because of complex inter-

actions between several user characteristics (Kumar

et al., 2018). To enable fake review detection Hus-

sain et al.(Hussain et al., 2020) implement a behav-

ioral method that utilizes thirteen different behavioral

features to calculate a review spam score for each re-

viewer. As review rankings are an important indicator

for the relevance of the reviews to demand-side partic-

ipants, the review consistency between review ranking

and review summary is crucial. To verify this consis-

tency Zhang et al.(Zhang et al., 2016) propose an ML

approach based on e-commerce data. This approach

enables demand-side participants to identify relevant

reviews based on the review rankings. As reviews are

often created by other customers, shopping assistance

is more responsive to the need of a special customer

and map the interaction with an employee digitally

and automated.

Bertacchini et al.(Bertacchini et al., 2017) devel-

oped for a robotic shopping assistant. These mea-

sures serve to promote sales in general and can guide

sales decisions. A shopping assistant can also be exe-

cuted on a customer’s smartphone as companion app

(Wulfert et al., 2019). Many online shops offer a full

24-hour service implementing lingual shopping as-

sistants. This service requires a lot of money when

done manually. Chatbots can be used as a solution

for automatic online shopping. Then the bot has to

be able to give an accurate and quick answer. For ex-

ample, Nursetyo et al.(Nursetyo et al., 2018) propose

an intelligent chatbot system that can be used as an

e-commerce assistant and therefore supports the pro-

cess of serving customers.

4.2.3 Handing-out Goods

The business task hand out goods includes the trans-

action in which the product becomes the customer’s

property. This process can be improved or auto-

mated with ML algorithms. Kourouthanassis and

Roussos(Kourouthanassis and Roussos, 2003) devel-

oped an unmanned and automated checkout system

with a smart shopping cart. Shopping carts are un-

usual in the fashion industry segment, for example,

Hauser et al.(Hauser et al., 2019) developed an ML

algorithm to determine whether a product has passed

the store exit or was only registered because it is

placed near the gate-mounted antennas. Suponenkovs

et al.(Suponenkovs et al., 2017) develop a system for

automatic invoice recognition by analyzing the pixels

with respect to their relevance and texture analysis.

As an ofﬂine scenario, payment of fruits or vegetables

in retail stores normally requires them to be manually

identiﬁed. Rojas et al.(Rojas-Aranda et al., 2020) or

Femling et al.(Femling et al., 2018) presents an image

classiﬁcation method with the goal of speeding up the

checkout process in stores.

Since incorrect operation or failures by employ-

ees at the POS accounts for 24% of inventory dif-

ferences in the retail industry (EHI, 2020), ML algo-

rithms are used to reduce these differences. Fraud

detection is assigned to the business task hand out

goods, since problems with the transaction in which

the product becomes the customer’s property should

be avoided. For interim transaction fraud detection

at the checkout, Trinh et al.(Trinh et al., 2011) devel-

oped a method that uses image recognition to deter-

mine the hand movements of the cashier. It should

be recognized by the algorithm whether all articles

that are handed out to the customer are also recorded

in a sales process, i.e. on the receipt. Fraud de-

tection in retrospect must be dealt with in a hu-

man resources management process. Pehlivanli et

al.(Pehlivanli et al., 2019) pursue a different approach

by analyzing transaction data. ML algorithms are

trained to classify fraud and non-fraud on the basis of

indicators like proﬁtability, stock turnover, stock cost,

and shelf life. Furthermore, ML is used to classify

the credit scores of customers in e-commerce. Kulka-

rni and Dhage(Kulkarni and Dhage, 2019) use ML

and data mining techniques to integrate information

crawled from social media to protect retailers from

payment defaults.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

338

4.2.4 Ordering Goods

We identiﬁed out-of-stock-detection and supplier

analysis as application areas of ML for the ordering

goods tasks. Out-of-stock describes a shelf space

in brick-and-mortar stores that are no longer ﬁlled

by the intended article and is empty. Customer de-

mands can no longer be fulﬁlled. As a detected out-

of-stock triggers the replenishment process, this appli-

cation area is assigned to the business task of making

goods available. For example, Paolanti et al.(Paolanti

et al., 2017) implement a mobile robot using visual

and textual analysis to facilitate automatic detection

of Shelf Out of Stock situations. Once the replenish-

ment process is triggered, a supplier must be selected.

Kuo et al.(Kuo et al., 2010) consider the aspect of

sustainability of suppliers in their supplier selection

optimization. ML algorithms are used to optimize

the supplier selection problems regarding operational

performance and environmental issues regarding the

criteria corporate social responsibility, service, cost,

environment, quality, and delivery.

5 DISCUSSION

Our structured literature review and the subsequent

structuring of the ML application areas along the

retail-related task depicted in the shell model (Figure

1) leads to the development of a framework for deter-

mining possible ML applications in retail (Figure 3).

Our structured literature review on ML application ar-

eas in retail resulted in three major ﬁndings. First, a

closer analysis of the framework reveals that there is a

trend for supporting decision-oriented and economic-

operative tasks in retail with ML. Other crucial tasks

in retail such as master data management as well as

technical and administrative task are either not cov-

ered by ML applications or underrepresented in our

literature sample. On the one hand, the support and

automation of economically valuable tasks seem rea-

sonable as they resemble the main bridging functions,

are the major value proposition, can be used to differ-

entiate from competitors, and are a major source of

revenue for retailers (Sch

utte, 2017). On the other

hand, master data and technological facilities con-

sist of the necessary data required for ML algorithms

(Weber and Sch

utte, 2019). If this data is inconsistent

or even not available, ML models provide false re-

sults or even cannot be trained. Second, our integra-

tive approach for the review in brick-and-mortar re-

tail and e-commerce is relatively balanced regarding

the application environment with 107 papers from of-

ﬂine and 118 papers from online retail. It also reveals

that ML applications make use of the peculiarities of

each environment. While the selection of an optimal

site for a brick-and-mortar store or the support of un-

manned checkout systems with ML is an ofﬂine case,

ML-based analyses of customers and product reviews

are only possible in e-commerce scenarios in which

these data exist (Hussain et al., 2020). As mentioned

earlier, e-commerce involves much data regarding its

customers and unstructured data regarding their prod-

ucts. This process is possibly further driven in future

by VR (virtual reality) shopping. There, besides click

and mouse movements in e-commerce, also hand,

eye, head, or leg movements could be tracked (Xi

and Hamari, 2019). We assume that existing appli-

cation areas presented in this work, especially from

e-commerce, can be adapted to VR shopping, and fu-

ture research may ﬁnd new application areas, espe-

cially for VR shopping (Xi and Hamari, 2019).

However, we also identiﬁed application areas that

are applied in both environments such as customer

conversion, promotion optimization, or product clas-

siﬁcation (Figure 4a). This trend is also supported

by the fact that brick-and-mortar retailers are increas-

ingly integrating technology within their stores to

bring in online convenience and experience (Dekimpe

et al., 2020).Product placement is a task necessary to

optimize revenues in brick-and-mortar stores but can

also be used in electronic shops to optimize the place-

ment of a product on a website (H

utsch and Wulfert,

2021). As these ML models can be used in both en-

vironments, the effort for implementing can be bal-

anced out by the dual application. Thus, these appli-

cations are useful for retails applying a multi- or om-

nichannel approach (Verhoef et al., 2015). Third, the

analysis of the application areas for the object of in-

terest reveals a tendency towards customer-centered

application of ML in e-commerce (40 papers) while

the focus in the ofﬂine environment is on a single or

set of articles (90 papers) (Figure 4b). Our analysis

also conﬁrms the developments described by Grewal

et al. (Grewal et al., 2017) towards a more customer-

centered focus in brick-and-mortar retail (28 papers).

However, we propose to focus even more on the cus-

tomer for a proper ML implementation and address

the customers’ experience for value generation (Ver-

hoef et al., 2009).For selecting the optimal site, it

is necessary to identify a location with an adequate

number of potential customers and select the right

articles for the needs of a speciﬁc customer milieu

(Wang et al., 2018).

Additionally, we concentrated on retail as the do-

main of application. Retail can be distinguished

from wholesale by treating small units, trading net

prices, and operating in a business-to-customer busi-

A Structured Literature Review on the Application of Machine Learning in Retail

339

(a) Ofﬂine vs. Online ML Application Areas.

Offline

Online

Offline

Online

Article

Customer

0 20 40 60 80 100

(b) Article- vs. Customer-centric ML Application Areas.

Figure 4: Application Area Analysis.

ness model (Levy et al., 2019). Although our frame-

work for ML application is particularly developed

for retail, we propose that many of the ML applica-

tions can also be used in a wholesale context as the

tasks are similar on a broad level (i.e., bridging func-

tions) comparing retail and wholesale architectures

(Sch

utte, 2017). We intentionally focused our liter-

ature research on ML applications in scientiﬁc pub-

lications to ensure a maximum level of quality and

comprehensibility of our research. However, we are

aware that implementations in practice can be ahead

of scientiﬁc publications (Eyes, 2021). Thus, we aim

to integrate practitioner’s sources such as company

white papers and publications of retail interest groups

in future research. We also opted for a broad research

scope reﬂected in our search string and the databases

queried. However, we did not provide a more detailed

analysis of the algorithms and data used in each ML

application with this broad scope. An important av-

enue for future research is to detail this overview and

provide evidence for proper algorithms to be imple-

mented within each application area. As we focused

on identifying ML applications in retail, we used a

task-oriented architecture by Sch

utte (Sch

utte, 2011)

that focuses on the business layer as a starting point.

However, we are aware that there exist other refer-

ence architectures covering additional application and

technical architecture layers in general and integrat-

ing ML in particular (Aulkemeier et al., 2016b). Al-

though these architectures include additional layers,

they are less detailed with regard to retail-speciﬁc

tasks. So it might be worthwhile for future research to

extend existing task-oriented architectures (Sch

utte,

2017; Becker and Sch

utte, 2004) with technical lay-

ers integrating ML. Alternatively, an architecture may

proﬁt from providing interfaces to plug in additional

services (e.g., ML services) making the architecture

more ﬂexible (Aulkemeier et al., 2016a).

6 CONCLUSION

We have identiﬁed 20 application areas of ML in of-

ﬂine and online retail based on a thorough analysis

of the current body of literature. The application

areas identiﬁed cover decision-oriented (business in-

telligence) and economic-operative tasks (managing

goods, serving customers, handing out goods, order-

ing goods). Following our analysis, ML can be im-

plemented to support and automatize structured and

unstructured tasks in retail. Additionally, current re-

search is equally concerned with the application of

ML in ofﬂine and online retail. For e-commerce, we

identiﬁed a tendency for a customer-centric usage of

ML, while in the brick-and-mortar context the arti-

cle is more often the object of interest. The contribu-

tion of our paper for practitioners and researchers is

a general overview of current research on ML appli-

cations in retail. For practitioners, the framework of

ML application can be applied by a retailer to check

for which tasks ML can be implemented in the com-

pany. Thus, it can be used as an indicator for the use-

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

340

fulness of an ML application to check whether an ML

implementation project is feasible and makes sense

from an economic perspective. The algorithms de-

veloped and applied in the identiﬁed papers can also

serve as a starting point for practical considerations

for implementation. For researchers, this work pro-

vides a retail-speciﬁc framework for ML application

in retail to foster the development of more holistic and

integrated ML models (e.g., promotion-sensitive re-

source optimization).

REFERENCES

Aulkemeier, F., Paramartha, M. A., Iacob, M. E., and van

Hillegersberg, J. (2016a). A pluggable service plat-

form architecture for e-commerce. Information Sys-

tems and e-Business Management, 14(3):469–489.

Aulkemeier, F., Schramm, M., Iacob, M.-E., and van

Hillegersberg, J. (2016b). A Service-Oriented E-

Commerce Reference Architecture. Journal of the-

oretical and applied electronic commerce research,

11:26–45.

Bandara, W., Furtmueller, E., Gorbacheva, E., Miskon, S.,

and Beekhuyzen, J. (2015). Achieving rigor in liter-

ature reviews: Insights from qualitative data analysis

and tool-support. Communications of the Association

for Information Systems, 37:154–204.

Barth, K., Hartmann, M., and Schr

oder, H. (2015). Be-

triebswirtschaftslehre des Handels, volume 7.,

uberar.

Springer Gabler, Wiesbaden.

Becker, J. and Sch

utte, R. (2004). Handelsinformation-

ssysteme : dom

anenorientierte Einf

uhrung in die

Wirtschaftsinformatik. Redline Wirtschaft, Frankfurt

am Main, 2., vollst edition.

Bertacchini, F., Bilotta, E., and Pantano, P. (2017). Shop-

ping with a robotic companion. Computers in Human

Behavior, 77:382–395.

Biem, A. (2003). A model selection criterion for clas-

siﬁcation: Application to hmm topology optimiza-

tion. Proceedings of the International Conference on

Document Analysis and Recognition, ICDAR, 2003-

Janua:104–108.

Blank, S. and Dorf, B. (2012). The startup owner’s manual:

The step-by-step guide for building a great company.

John Wiley & Sons, Hoboken, New Jersey.

Bousqaoui, H., Achchab, S., and Tikito, K. (2018). Ma-

chine learning applications in supply chains: An em-

phasis on neural network applications. In Proceedings

of 2017 International Conference of Cloud Computing

Technologies and Applications, CloudTech 2017, vol-

ume 2018-January, pages 1–7. Institute of Electrical

and Electronics Engineers Inc.

Chandrashekhara, K., Thungamani, M., Babu, C. G., and

Manjunath, T. (2019). Smartphone price prediction in

retail industry using machine learning techniques. In

Emerging Research in Electronics, Computer Science

and Technology, pages 363–373. Springer.

Chopra, S., Meindl, P., and Kalra, D. V. (2013). Supply

chain management: strategy, planning, and operation,

volume 232. Pearson Boston, MA.

Cooper, H. M. (1988). Organizing Knowledge Synthesis:

A Taxonomy of Literature Reviews. Knowledge in

Society, 1(1):104–126.

Dekimpe, M. G., Geyskens, I., and Gielens, K. (2020). Us-

ing technology to bring online convenience to ofﬂine

shopping. Marketing Letters, 31(1):25–29.

D’Haen, J., Van Den Poel, D., and Thorleuchter, D. (2012).

Predicting customer proﬁtability during acquisition:

Finding the optimal combination of data source and

data mining technique. Expert Systems with Applica-

tions, 40(6):2007–2012.

EHI (2020). Inventurdifferenzen 2020.

Eyes, W. (2021). Welcome to the visual intelligence era.

Femling, F., Olsson, A., and Alonso-Fernandez, F. (2018).

Fruit and vegetable identiﬁcation using machine

learning for retail applications. In 2018 14th Inter-

national Conference on Signal-Image Technology &

Internet-Based Systems (SITIS), pages 9–15. IEEE.

Ferstl, O. K. and Sinz, E. J. (2013). Grundlagen der

Wirtschaftsinformatik, volume 7., aktual. Oldenbourg

Verlag, M

unchen.

Feurer, M., Klein, A., Eggensperger, K., Springenberg,

J. T., Blum, M., and Hutter, F. (2019). The springer

series on challenges in machine learning. Nips, pages

113–134.

Frey, C. B. and Osborne, M. A. (2017). THE FUTURE OF

EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS

TO COMPUTERISATION? Technological forecast-

ing and social change, 114:254–280.

Goos, M., Manning, A., and Salomons, A. (2009). Job

Polarization in Europe. American Economic Review,

99(2):58–63.

Grewal, D., Roggeveen, A. L., and Nordf

allt, J. (2017). The

Future of Retailing. Journal of Retailing, 93(1):1–6.

Große Holtforth, D. (2018). Machine Learning im E-

commerce.

Hauser, M., G

unther, S. A., Flath, C. M., and Thiesse, F.

(2019). Towards digital transformation in fashion re-

tailing: A design-oriented is research study of auto-

mated checkout systems. Business and Information

Systems Engineering, 61(1):51–66.

Heinrich, L. and Stelzer, D. (2009). Informationsmanage-

ment: Grundlagen, Aufgaben, Methoden. Oldenbourg

Wissenschaftsverlag, M

unchen.

Henzel, J. and Sikora, M. (2020). Gradient boosting and

deep learning models approach to forecasting promo-

tions efﬁciency in fmcg retail. In International Con-

ference on Artiﬁcial Intelligence and Soft Computing,

pages 336–345. Springer.

Huber, J. and Stuckenschmidt, H. (2020). Daily retail de-

mand forecasting using machine learning with empha-

sis on calendric special days. International Journal of

Forecasting.

Hussain, N., Turab Mirza, H., Hussain, I., Iqbal, F., and

Memon, I. (2020). Spam Review Detection Using the

Linguistic and Spammer Behavioral Methods. IEEE

Access, 8:53801–53816.

A Structured Literature Review on the Application of Machine Learning in Retail

341

utsch, M. and Wulfert, T. (2021). Implementation of

Placement Optimization in Electronic Commerce: To-

wards a Data Generation Method to Increase Forecast

Elasticity. In Proceedings of the 25th Paciﬁc Asia

Conference on Information Systems (PACIS2021),

pages 1–12, Dubai.

Jheng, Y. R. and Luo, Z. C. (2019). Customer Retention

Prediction with CNN. In Proceedings of the 4th In-

ternational Conference on Data Mining and Big Data

(DMBD 2019), volume 1071, pages 104–113, Chiang

Mai, Thailand. Springer.

Katarya, R. and Verma, O. P. (2017). Effectual rec-

ommendations using artiﬁcial algae algorithm and

fuzzy c-mean. Swarm and Evolutionary Computation,

36(June 2016):52–61.

Khatwani, G. and Srivastava, P. R. (2016). Real-time pre-

diction of information search channel using data min-

ing techniques. In Proceedings of the 2015 Interna-

tional Conference on Green Computing and Internet

of Things, ICGCIoT 2015, pages 924–929.

Kim, K. and Lee, J. (2012). Sequential manifold learning

for efﬁcient churn prediction. Expert Systems with Ap-

plications, 39(18):13328–13337.

Kourouthanassis, P. and Roussos, G. (2003). Developing

consumer-friendly pervasive retail systems. IEEE Per-

vasive Computing, 2(2):32–39.

Kulkarni, S. V. and Dhage, S. N. (2019). Advanced credit

score calculation using social media and machine

learning. Journal of Intelligent and Fuzzy Systems,

36(3):2373–2380.

Kumar, A., Kabra, G., Mussada, E. K., Dash, M. K., and

Rana, P. S. (2019). Combined artiﬁcial bee colony al-

gorithm and machine learning techniques for predic-

tion of online consumer repurchase intention. Neural

Computing and Applications, 31:877–890.

Kumar, D. T., Soleimani, H., and Kannan, G. (2014).

Forecasting return products in an integrated for-

ward/reverse supply chain utilizing an anﬁs. Interna-

tional Journal of Applied Mathematics and Computer

Science, 24(3):669–682.

Kumar, N., Venugopal, D., Qiu, L., and Kumar, S. (2018).

Detecting Review Manipulation on Online Platforms

with Hierarchical Supervised Learning. Journal of

Management Information Systems, 35(1):350–380.

Kuo, R. J., Wang, Y. C., and Tien, F. C. (2010). Integra-

tion of artiﬁcial neural network and mada methods for

green supplier selection. Journal of Cleaner Produc-

tion, 18(12):1161–1170.

Laudon, K. C. and Traver, C. G. (2020). E-Commerce 2018:

business. technology. society, volume 15. Pearson Ed-

ucation, Boston.

Levy, M., Weitz, B. A., and Grewal, D. (2019). Retail-

ing management. McGraw-Hill Education, New York,

NY, tenth edit edition.

Li, X. and Li, Z. (2019). A hybrid prediction model for

e-commerce customer churn based on logistic regres-

sion and extreme gradient boosting algorithm. Inge-

nierie des Systemes d’Information, 24(5):525–530.

Liang, J., Ramanauskas, B., and Kurenkov, A. (2019). Job

Loss Due To AI — How Bad Is It Going To Be?

Liang, R. and Wang, J. q. (2019). A Linguistic Intuition-

istic Cloud Decision Support Model with Sentiment

Analysis for Product Selection in E-commerce. Inter-

national Journal of Fuzzy Systems, 21(3):963–977.

Loureiro, A. L., Migu

eis, V. L., and da Silva, L. F. (2018).

Exploring the use of deep neural networks for sales

forecasting in fashion retail. Decision Support Sys-

tems, 114:81–93.

Machado, M. A. S., Gomes, L. F. A. M., Santos, D. J., and

Caldeira, A. M. (2015). Using a bipolar choquet neu-

ral network to locate a retail store. Procedia Computer

Science, 55:741–747.

Mahaboob Basha, A. M., Rajaiah, M., Penchalaiah, P.,

Raja Kamal, C. H., and Niranjana Rao, B. (2020).

Machine learning-structural equation modeling algo-

rithm: The moderating role of loyalty on customer re-

tention towards online shopping. International Jour-

nal of Emerging Trends in Engineering Research,

8(5):1578–1585.

Manyika, J., Chui, M., Miremadi, M., Bughin, J., George,

K., Willmot, P., and Dewhurst, M. (2017). A Future

That Works: Automation, Employment, and Produc-

tivity.

McAfee, A. and Brynjolfsson, E. (2017). Machine, Plat-

form, Crowd: Harnessing Our Digital Future. W W

Norton & Co Inc, New York City LB - Mcafee2017.

Miao, Y. (2020). A machine-learning based store layout

strategy in shopping mall. In International Conference

on Machine Learning and Big Data Analytics for IoT

Security and Privacy, pages 170–176. Springer.

Mitchell, T. M. et al. (1997). Machine learning. McGraw-

Hill.

Nicola, M., Alsaﬁ, Z., Sohrabi, C., Kerwan, A., Al-Jabir,

A., Iosiﬁdis, C., Agha, M., and Agha, R. (2020). The

socio-economic implications of the coronavirus pan-

demic (COVID-19): A review. International Journal

of Surgery, 78(1):185–193.

Nikulin, V. (2016). Prediction of the shoppers loyalty with

aggregated data streams. Journal of Artiﬁcial Intelli-

gence and Soft Computing Research, 6(2):69–79.

Niu, X., Li, C., and Yu, X. (2017). Predictive analytics of

E-commerce search behavior for conversion. In Pro-

ceedings of the America’s Conference on Information

Systems (AMCIS 2017), pages 1–10.

Nursetyo, A., Subhiyakto, E. R., et al. (2018). Smart chat-

bot system for e-commerce assitance based on aiml.

In 2018 International Seminar on Research of Infor-

mation Technology and Intelligent Systems (ISRITI),

pages 641–645. IEEE.

Paolanti, M., Sturari, M., Mancini, A., Zingaretti, P., and

Frontoni, E. (2017). Mobile robot for retail survey-

ing and inventory using visual and textual analysis of

monocular pictures based on deep learning. In 2017

European Conference on Mobile Robots (ECMR),

pages 1–6. IEEE.

Pehlivanli, D., Eken, S., and Ayan, E. (2019). Detection

of fraud risks in retailing sector using mlp and svm

techniques. Turkish Journal of Electrical Engineering

& Computer Sciences, 27(5):3633–3647.

ICEIS 2022 - 24th International Conference on Enterprise Information Systems

342

Randolph, J. J. (2009). A guide to writing the dissertation

literature review. Practical Assessment, Research and

Evaluation, 14(13):1–13.

Rojas-Aranda, J. L., Nunez-Varela, J. I., Cuevas-Tello, J. C.,

and Rangel-Ramirez, G. (2020). Fruit classiﬁcation

for retail stores using deep learning. In Mexican Con-

ference on Pattern Recognition, pages 3–13. Springer.

Rudolph, T., Nagengast, L., Melanie, B., and Bouteiller,

D. (2015). Die Nutzung mobiler Shopping Apps im

Kaufprozess. Marketing Review St. Gallen, 3(32):42–

49.

Sch

utte, R. (2011). Modellierung von handelsinfor-

mationssystemen. Kumulative Habilitationsschrift.

Westf

alische Wilhelms-Universit

at M

unster. M

unster.

Sch

utte, R. (2017). Information systems for retail compa-

nies. In Dubois, E. and Pohl, K., editors, International

Conference on Advanced Information Systems Engi-

neering, pages 13–25, Essen, Deutschland. Springer.

Sch

utte, R. and Vetter, T. (2017). Analyse des digital-

isierungspotentials von handelsunternehmen. In Han-

del 4.0, pages 75–113. Springer.

Singh, P. and Agrawal, V. (2019). A Collaborative Model

for Customer Retention on User Service Experience.

In Bhatia, S. K., Shailesh, T., Mishra, K. K., and Ti-

wari, M. C., editors, Advances in Computer Commu-

nication and Computational Sciences, pages 55–64.

Springer Nature, Singapore, 924 edition.

Suponenkovs, A., Sisojevs, A., Mos

ans, G., Kampars, J.,

Pinka, K., Grabis, J., Locmelis, A., and Taranovs, R.

(2017). Application of image recognition and machine

learning technologies for payment data processing re-

view and challenges. In 5th IEEE Workshop on Ad-

vances in Information, Electronic and Electrical En-

gineering (AIEEE), pages 1–6.

Thornton, C., Hutter, F., Hoos, H. H., and Leyton-Brown,

K. (2013). Auto-weka: Combined selection and

hyperparameter optimization of classiﬁcation algo-

rithms. Proceedings of the ACM SIGKDD Interna-

tional Conference on Knowledge Discovery and Data

Mining, Part F1288:847–855.

Trinh, H., Fan, Q., Jiyan, P., Gabbur, P., Miyazawa, S., and

Pankanti, S. (2011). Detecting human activities in

retail surveillance using hierarchical ﬁnite state ma-

chine. In ICASSP, IEEE International Conference on

Acoustics, Speech and Signal Processing - Proceed-

ings, pages 1337–1340.

Veres, S. M., Molnar, L., Lincoln, N. K., and Morice, C. P.

(2011). Autonomous vehicle control systems - A re-

view of decision making. Proceedings of the Institu-

tion of Mechanical Engineers. Part I: Journal of Sys-

tems and Control Engineering, 225(3):155–195.

Verhoef, P. C., Kannan, P. K., and Inman, J. J. (2015). From

Multi-Channel Retailing to Omni-Channel Retailing.

Introduction to the Special Issue on Multi-Channel

Retailing. Journal of Retailing, 91(2):174–181.

Verhoef, P. C., Lemon, K. N., Parasuraman, A., Roggeveen,

A., Tsiros, M., and Schlesinger, L. A. (2009). Cus-

tomer Experience Creation: Determinants, Dynam-

ics and Management Strategies. Journal of Retailing,

85(1):31–41.

Verstraete, G., Aghezzaf, E. H., and Desmet, B. (2019). A

data-driven framework for predicting weather impact

on high-volume low-margin retail products. Journal

of Retailing and Consumer Services, 48:169–177.

Vinodhini, G. and Chandrasekaran, R. M. (2014). Opinion

mining using principal component analysis based en-

semble model for e-commerce application. CSI Trans-

actions on ICT, 2(3):169–179.

Vom Brocke, J., Simons, A., Niehaves, B., Riemer, K., Plat-

tfaut, R., Cleven, A., Brocke, J. V., and Reimer, K.

(2009). Reconstructing the Giant: On the Importance

of Rigour in Documenting the Literature Search Pro-

cess. In Proceedings of the 17th European Conference

on Information Systems (ECIS ’17), volume 9, page

2206–2217, Verona, Italy.

Wan, M., Ni, J., Misra, R., and McAuley, J. (2020). Ad-

dressing marketing bias in product recommendations.

In WSDM 2020 - Proceedings of the 13th Interna-

tional Conference on Web Search and Data Mining,

pages 618–626.

Wang, L., Fan, H., and Wang, Y. (2018). Site selection of

retail shops based on spatial accessibility and hybrid

bp neural network. ISPRS International Journal of

Geo-Information, 7(6):202.

Weber, F. and Sch

utte, R. (2019). State-of-the-art and

adoption of artiﬁcial intelligence in retailing big-

data-analysen im promotion management view project

coming closer to reality-neural correlates of tam view

project digital policy, regulation and governance state-

of-the-art and adop. researchgate.net.

Weber, F. D. and Sch

utte, R. (2019). State-of-the-art and

adoption of artiﬁcial intelligence in retailing. Digital

Policy, Regulation and Governance.

Wulfert, T., Betzing, J. H., and Becker, J. (2019). Elic-

iting Customer Preferences for Shopping Companion

Apps: A Service Quality Approach. In Proceedings of

the 14th Internationale Tagung Wirtschaftsinformatik

(WI2019), pages 1220–1234, Siegen, Deutschland.

Xi, N. and Hamari, J. (2019). Vr shopping: A review of

literature. In AMCIS 2019 Proceedings.

Yu, W., Sun, Z., Liu, H., Li, Z., and Zheng, Z. (2018).

Multi-level deep learning based e-commerce product

categorization. In eCOM@ SIGIR.

Zhang, D., Zhou, L., Kehoe, J. L., and Kilic, I. Y. (2016).

What Online Reviewer Behaviors Really Matter? Ef-

fects of Verbal and Nonverbal Behaviors on Detection

of Fake Online Reviews. Journal of Management In-

formation Systems, 33(2):456–481.

A Structured Literature Review on the Application of Machine Learning in Retail

343