A Proof-of-Concept Web-Based Course Recommender System Using a Hybrid Model of Content-Based Filtering and Collaborative Filtering

Abstract

This thesis presents a proof-of-concept web-based course recommender system for personalized learning. The main aim is to design and implement a practical recommendation system that can suggest suitable courses based on learner interests and past interaction history. The system uses a hybrid model that combines Content-Based Filtering and Collaborative Filtering.

The Content-Based Filtering component analyzes course information using TF-IDF and cosine similarity to find courses that are semantically relevant to the learner’s preferences, while the Collaborative Filtering component uses Singular Value Decomposition to learn patterns from community rating behavior and predict user preference for unseen courses. These two methods are combined in a hybrid recommendation model to improve recommendation quality by balancing content relevance and community-based prediction.

The system is implemented as a proof-of-concept web-based prototype. Python, Scikit-learn, Surprise, and FastAPI are used to build the recommendation engine and expose its functionality as a simple web-based prototype. The model is evaluated using a real public Coursera dataset, and the results show that the hybrid approach provides more balanced recommendation performance than the individual methods alone. Overall, the study demonstrates that a hybrid recommender model can serve as a practical foundation for personalized course recommendation in e-learning systems.