COMP5425
Week9 Semester 1, 2025
Recommender Systems
Background
Recommendation algorithms
Collaborative filtering
◼ User based
◼ Model based
◼ Matrix factorization
Content-based
◼ Product, document, image, video, audio
Learning based
Context Aware Recommendation
Evaluation
Recommendation is everywhere
eCommerce
Amazon, eBay, …
Social
Facebook, LinkedIn, …
◼ Friends, groups, jobs
Media
Youtube, Netflix, Spotify, …
News
Advertisement
Others
MOOC, tourism, …
Benefits of RecSys
For customers or users
Find relevant things
Narrow down the set of choices
Help explore the space of options
Discover new things
…
For providers or vendors
Additional and probably unique personalized or customized service
Increase trust and customer loyalty
Increase sales (30% - 70%), click through rates, conversion etc.
Opportunities for promotion, persuasion
Obtain more knowledge about customers
…
Problem Statement
Input
User model and profile (e.g., ratings, preferences, and other meta. data)
Items (with or without attributes)
Goal
Recommend items to potential users
◼ Relevance score in terms of various criteria (e.g., context)
Obtain missing values between users and items
◼ Netflix: 100K movies, 10M users, 1B ratings