ADDIS ABABA SCIENCE AND TECHNOLOGY UNIVERSITY
Topic - An Intelligent Recommendation System for E-commerce Systems
based on User Activity
Department of Software Engineering
February, 2023
Individual Research Presentation
Contents
• Background
• Target Audience
• Problem Statement
• Proposed Solution
• Research Questions
• Literature Review
• Research Gap
• Objectives
• Research Methodology
• Benefits
• Action Plan
01 Background
Electronic commerce has become more prevalent and is on the rise as a result of the
development of technology in the tools and services made available online as well as the
accessibility of contemporary hardware and software.
Due to the abundance of products offered, clients frequently experience product overload and
have trouble selecting the best option. Because of the increased competition between
commercial cities as a result of these variables, it is necessary to work more effectively in order
to improve financial earnings. By helping customers identify the right products based on their
interests, recommendation systems seek to increase the effectiveness of the e-commerce
system.
02 Target Audience
The system is aimed to greatly beneficial to Small and Medium sized Enterprises (
SMEs) who have chosen to put out their products online.
EthioSuQ, Sheger, Qefira, Helloomart, Sheba Shopping, Guzomart, Kikuu Ethiopia, Kemmis,
Merkatoonline, Shola Mart, and many more providing an online shopping platform with a
selection of leather products, coffee, personal care, apparel, jewelry, sporting goods,
shoes, hardware & tools, furniture, homemade seasonings.
03 Statement of the Problem
There are challenges associated with sparsity, diversity, and scalability. More
sophisticated ecommerce websites are required as more firms move their operations
online due to the competition to draw visitors and keep them on the site longer to
increase income.
Due to the pressure on industries to expand their consumer base into the Ecommerce
platform, there is a demand for a smart system that, on the one hand, aids the industry in
generating more revenue and gaining a firm grip on their customers, while, on the other
hand, aids consumers by assisting them in making a personalized selection from the vast
array of options and presenting products they are most likely to purchase.
04 Proposed Solution
A Hybrid recommendation system based on collaborative filtering (CF) algorithm helped
by user activity data collected throughout the user’s interaction with the system. The
proposed recommender system is developed based on the customers’ navigational
activities in e-commerce sites including browsing, searching, product click, cart
placement. This solution aims to combine the results of CF and activity analysis to solve
the common problems with CF.
05 Research Questions
1. How can recommender systems leverage hybrid data—the navigational information
about a client that is currently available—to produce the best product recommendations?
2. How can recommender systems predict how accurate their suggestions will be in
advance?
3. How does the suggested solution address the primary drawbacks of the earlier
recommendation systems?
06 Objectives
The main objective of the study is to propose an intelligent recommendation system which
combines Collaborative Filtering (CF) algorithms and user navigational activity data to
produce a hybrid result.
Specific Objectives
1. To employ hybrid data, i.e., system user data and CF data.
2. To demonstrate that the proposed system's recommendations are more accurate than those from
earlier systems.
3. To compare the new system's performance empirically to the difficulties that the earlier algorithms
had to overcome.
07 Literature Review
The first recommender system was created twenty years ago (Goldberg, Nichols, Oki,
& Terry, 1992).
Ethiopia , In the past couple of years, some popular e-commerce platforms have
emerged by the same business model to connect businesses to consumers and
sellers to buyers.
E-commerce is growing fast in Ethiopia but native recommendation systems are not
widely used.
Research Gap
The majority of the RS problems, but not all of them have been identified in the LR.
High time complexity due to the many computations that must be made, which may get
more difficult with system scaling.
Cold-start - refers to the difficulty in finding sufficient reliable similar users.
Sparsity - refers to the difficulty in generating accurate recommendations for the cold
users who only rated a small number of items.
08 Research Methodology
The product preferences matrix, a global matrix including all the items and containing the
computations for the features for each product, is built using these features in tandem with being
used to categorize the preferences for the products. This matrix is maintained in the products
database and updated if the features change.
The customer preference matrix is retrieved for each user who logs in to the system using their unique
identity, and it is updated if there are any changes to the user's preferences.
To create suggestion lists, a variety of entities are employed, including:
1. Consumer activity 3. A matrix of customer preference.
2. Product attributes 4. A matrix of product features.
09 BENEFITS
1. Increased User Engagement
User engagement is at the heart of every eCommerce
solution. When recommendation engines provide users’
suggestions based on their preferences user love the
website.
2. Increased Sales and Revenue
SMEs can use recommendation engines to tackle
eCommerce challenges, leading to more revenue and
increased sales.
Cont…
1. Decreased Cold Start
The suggested approach uses statistical analysis to provide a
recommendation based on the goods' preference matrices in
order to address the issue of a cold-start.
2. Better Tailored Recommendations
The suggested system will provide a list more tailored to
the user and their preferences.
10 HOW IT WORKS
First, the user-based CF methods recommend items by taking into account the
feedback (ratings) of users with similar preferences to the target user.
Secondly, the CBF approaches operate with the similarity of the items, so similar
items to the ones liked by the target user are recommended.
Finally, . The customer preference matrix is retrieved for each user who logs in to the
system using their unique identity, and it is updated if there are any changes to the
user's preferences.