Inquiry Process Document:

Contribution of machine learning on online shopping platforms in predicting customer

purchasing decisions
Overall word count: 1462 words

Section A: Inquiry Focus

Section A Word Count: 292 words

Inquiry Question:
To what extent does machine learning used on online shopping platforms such as Shopee
contribute to accuracy and efficiency of predicting customer purchasing decisions?

Connection:
In Malaysia, over 145,000 local Shopee sellers achieved 25% growth in year-on-year sales on
the marketplace due to superior levels of customer service and optimised storefronts from
October last year to September 2022, according to the Shopee Super Awards 2022 Milestone
study (2022). These changes are due to the implementation of machine learning across the
e-commerce platform. Machine learning can analyse different aspects of users on the platform
to generate an output where they can predict purchasing decisions. Even the design of the
Shopee platform has already been revised in order to influence and impact the purchasing
behaviour of customers. With the myriad of data extracted from multiple sources, they can
predict whether a customer will buy a certain product or not.

Context: 4.2 Economic

The context that can be connected with the real world example is economic, since knowledge of
factors affecting users' behaviour in purchasing will help online shopping platforms improve their
sites and increase their revenue. This will contribute to the economic growth of the platform.

Content: 3.1 Data & 3.6 Artificial Intelligence

The content associated with the inquiry is data since throughout the document, there will be a
lot of discussion on the data usage of platform users. The information of a user is obtained and
analysed with machine learning in order to provide insight. This brings us to the next content
which is artificial intelligence since machine learning is used to predict the purchasing decisions
of a user.

Concept: 2.6 Systems & 2.2 Expression

The concept that can be linked to the real world example is systems, as the users of online
shops interact with the systems on the platforms to browse a product. Other than that,
expression is also a concept that can be linked to real life examples since the behaviour of a
customer can be analysed using machine learning.
Section B: Claims and Perspective
Section B Word Count: 1160 words
Buyer-seller interactions have evolved as a result of e-commerce. In an effort to comprehend
and forecast customer purchasing intent, this has led to the development of successful
personalised offerings and improved platforms. In particular, this section will cover the
application of machine learning to forecast consumer behaviour when making purchases.

Source A (395 words)

Stubseid, S., & Arandjelović, O. (2018). Machine Learning Based Prediction of Consumer
Purchasing Decisions: The Evidence and Its Significance.
https://research-repository.st-andrews.ac.uk/bitstream/handle/10023/13634/Stubseid_Machine_l
earning_AAAI_18_AAM.pdf?sequence=1&isAllowed=y

The study discussed easily quantifiable elements that affect consumers' decisions to buy. The
researchers assess and evaluate how well various models predict consumer behaviour using
data-driven machine learning techniques. The easily measured components can be divided into
two categories: variables specific to the consumer and variables linked to the product. In terms
of consumer-specific variables, these comprise contextual factors (such as the consumer's
financial status, level of debt, etc.) and demographic data about the customer (such as age, sex,
income, and education). Price, product category, discounts, and gifts are compromised when it
comes to product-related variables.

There are 642,709 entries in the data, each of which represents a specific client purchase
decision. The study's data set is referred to as the biggest in the published literature. The
representation of the two classes, which are those who make purchases and those who do not,
is balanced. This equilibrium makes sure that there is no bias in the analysis or predictions
favouring one class over the other. The performance of the two categorization strategies used in
the study which are Random Forest and Naïve Bayes was compared. In terms of average
classification accuracy, AUC, and F1-score, the random forest classifier performed better than
the Naïve Bayes classifier.

Furthermore, the study discovered that when a buyer makes a successful purchase, there is a
higher level of interaction and less independence across features. This demonstrates the way
various e-commerce platform features and elements interact when a client makes a purchase,
as opposed to just one feature helping customers make wise judgements.

The credibility of this source is assured since it was written by two individuals at a respectable
university called the University of St. Andrews. The writers even have ties to this university's
School of Computer Science. The large volume of data also indicates that accuracy will be
higher because the machine learning output will be able to identify more trends and patterns. A
wider view of consumer behaviour is provided by the huge and diversified data set used for the
analysis of classification algorithms.
Overall, I think the study offers insightful information about how to address my inquiry question
because it addresses consumer behaviour and the ways in which various customer
characteristics might influence a customer's decision to buy. Companies and consumers looking
to better understand and influence their purchase decisions should take note of these findings.

Source B (374 words)

Morsi, S. (2020). A Predictive Analytics Model for E-commerce Sales Transactions to Support
Decision Making: A Case Study. International Journal of Computer and Information
Technology(2279-0764), 9(1). https://doi.org/10.24203/ijcit.v9i1.3

The aim of this study is to illustrate the importance and applicability of predictive analytics
models for online retailers. Predictive analytics is applied to online transaction data to project
product sales and assess customer attributes, with the primary objective of assisting
e-commerce businesses in making better decisions. The paper focuses on the use of predictive
modelling frameworks and machine learning algorithms to forecast the total purchase amounts
of online customers based on their distinctive characteristics.

The dataset used in this study was historical online transaction data from an Egyptian
e-commerce company. In this paper, linear regression is the supervised machine learning
approach applied. It is especially effective in forecasting quantitative reactions, such as the total
amount of money that internet buyers would spend. The model predicts and minimises errors
based on the input features, iteratively adjusting parameters until the error is minimal enough.
The model's performance was estimated using the mean squared error (MSE) and mean
absolute error (MAE) methods from the 'sklearn' package.

The data indicates that men are more likely than women to make purchases online, with 76.8%
of online shoppers being men. Men's shopping activity peaked in the age group of 26 to 35
years old, accounting for around 45% of the results. Pair plots were used to display the data,
and descriptive analysis was used to look at purchasing distributions. It helps identify the
connections between the variables affecting sales. Experts in the e-commerce sector can learn
a lot from the data of consumer behaviour in relation to product category purchases. Sales
strategies that consider the interests and behaviour of the target market can be established with
the help of this data.

The paper was written by Shereen Morsi. She works in the Business Information Systems
Department of the Arab Academy for Science, Technology, & Maritime Transport in Cairo,
Egypt. The author's involvement with academic institutions serves as evidence of her
knowledge in the topic. This demonstrates how trustworthy the research is as a source for this
question.

This study is significant in addressing questions about the use of complex analytics, particularly
predictive algorithms, to estimate consumer purchasing habits by showing how predictive
modelling supports flexible decision-making. It also examined the influence of each customer's
distinct characteristics on prediction models.

Source C (341 words)

Liu, L. (2022). e-Commerce Personalized Recommendation Based on Machine Learning
Technology. Mobile Information Systems, 2022, 1–11. https://doi.org/10.1155/2022/1761579

The research paper explores about the use of machine learning techniques to create a
customised e-commerce recommendation system. The e-commerce personalised
recommendation system makes use of machine learning to map inputs to outputs, with a
particular emphasis on the synthesis and induction functions. By removing pertinent data to
forecast the target value, it aids in the construction of a training dataset. It also highlights the
importance of tailored suggestions in assisting clients. This will help customers to sift through
mountains of data, thus leading to wise choices.

This study discusses different recommender systems e-commerce use which is collaborative
filtering, content-based and user-based. User-based recommender systems provide
recommendations to consumers by examining the actions and behaviours of users who have
similar attributes to them. Then, the algorithm suggests products for its customers based on
what users with similar preferences have engaged with. Next, collaborative filtering
recommender systems make recommendations for products based on the preferences and
actions of a sizable user base. It creates individualised recommendations for a single customer
by spotting trends and parallels among users on a large scale. Lastly, recommender systems
that are content-based suggest products according to the qualities and features of the products
themselves. It recommends products based on similar qualities of previous searches or
purchases made by the user. The reason for evaluating these recommender systems is to
understand the way the algorithm delivers tailored recommendations for clients. The study then
comes to the conclusion that factors such as income levels, commodity prices, online shopping
experiences, product quality, suggestion relevancy, credit rating, service quality, and purchasing
intention all significantly increase the likelihood that people will shop.

The author of the study, Liping Liu, is regarded as reputable. His connection to Wuhan College's
School of Management in China adds to his legitimacy and highlights the importance of the
study on the application of machine learning.

In general, machine learning plays a critical role in the e-commerce personalised

recommendation system by offering users precise and instantaneous recommendations,
addressing problems like information overload, and improving the entire shopping experience.
List of References

Sources used for the inquiry (IPD, Multimedia presentation)

Secondary sources
Stubseid, S., & Arandjelović, O. (2018). Machine Learning Based Prediction of Consumer
Purchasing Decisions: The Evidence and Its Significance.
https://research-repository.st-andrews.ac.uk/bitstream/handle/10023/13634/Stubseid_Machine_l
earning_AAAI_18_AAM.pdf?sequence=1&isAllowed=y

Morsi, S. (2020). A Predictive Analytics Model for E-commerce Sales Transactions to Support
Decision Making: A Case Study. International Journal of Computer and Information
Technology(2279-0764), 9(1). https://doi.org/10.24203/ijcit.v9i1.3

Liu, L. (2022). e-Commerce Personalized Recommendation Based on Machine Learning

Technology. Mobile Information Systems, 2022, 1–11. https://doi.org/10.1155/2022/1761579

Primary sources
Google Form for Questionnaire
Interview with Wan Noraini - A Shopee user since 2020
Interview with Syasya Mardhiha - A Shopee user since 2020

Artificial Intelligence tools

ChatGPT: Refining video script, generate questions for interviewee, finding synonyms for words
used in the document
Quillbot: Checking IPD grammar

Sources used for constructing the multimedia presentation

Media sources
pexels.com

Applications /online tools

Canva
CapCut