The School of Computer Science at the University of Windsor is pleased to present …
MSc Thesis Proposal by: Emmanuel Ainoo
Date: Friday, October 6th, 2023
Time: 2:00 pm to 4:00 pm
Location: Lambton Tower Room 3105
E-commerce recommendation accuracy can be improved by transferring user preference insights from social media domains like Facebook to the e-commerce domain, especially when dealing with users who have limited historical purchase activity.
An existing cross domain recommendation system GaoLinRec23 system employs (CMF) Collective Matrix Factorization to jointly factorize user-item interaction matrices from both e-commerce and social media domains. The system first collects user-item interaction data from both domains, including user activity such as clicks and purchases. Next, CMF is applied to these user-item interaction matrices, factorizing these matrices to decompose them into underlying latent patterns. The system then uses the decomposed user-item interaction matrices, which now reveal the underlying latent patterns, as input data for collaborative filtering technique.
However, CMF, as used in the GaoLinRec23, and other existing systems ChenWuRec05, Gao2019Rec, WangHeNieChuaRec17, assumes that specific item data such as product details (e.g., [product1: skirt, product2: cell phone]) is common between selected domains, which may not reflect the practical reality. In practice, most e-commerce and social media domains often do not share item details across domains due to varying user behaviours and preferences. It then becomes a problem to apply existing system approaches to real-world platforms like Amazon, Facebook which do not share product/item data.
This thesis proposes Social Media Cross Domain Recommendation System ‘2023 (SM-CDRS ’23), which mines users’ social media activities such as likes, posts, comments, and e-commerce historical activity such as purchases and clicks to create a unified dataset (user keys, likes, comments, purchases) using the proposed MDUK (Merge Datasets by User Unique Key) algorithm. By using (GSP) Generalized Sequential Pattern Mining to discover sequential patterns, sequential rule mining is then used to establish association rules between user preferences across both domains. For instance, a rule like “Users (user) who typically like cooking posts on social media, purchase cooking recipes (item) from e-commerce”. With such rules, the system can unearth sequential patterns in user -item associations for rule-based collaborative filtering without having to share the item details across the domains, as existing systems do. A pattern such as <, (likes cooking posts(A)), (purchases cooking recipes(B))> means, whenever likes A, they follow it up with a purchase of B.
External Reader: Dr. Jeffrey Rau
Internal Reader: Dr. Shafaq Khan
Advisor: Dr. Christie Ezeife