Yousef Tarek Nathan — AI & Data Science Developer

AI & Machine Learning

Course Recommendation System

Personalized course recommendation system combining BERT embeddings, semantic similarity, PageRank, and hybrid scoring to recommend relevant academic courses.

TECHNOLOGIES

Python · BERT · NLP · Sentence Transformers · PageRank · Scikit-learn · Jupyter Notebook

PROBLEM

Recommend relevant courses to users by understanding semantic relationships between course content and modeling connections between courses.

APPROACH

Generate BERT-based embeddings, calculate semantic similarity, construct a course relationship graph, apply PageRank, and combine the resulting scores through a hybrid recommendation approach.

DATASET

2,000 Coursera courses across 11 academic categories with 58,784 course relationships.

IMPLEMENTATION

Implemented text preprocessing, BERT embeddings, semantic similarity, graph-based ranking with PageRank, Top-K k-NN optimization, hybrid scoring, and an interactive recommendation interface.

RESULTS

Processed 2,000 courses and 58,784 course relationships, with Top-K optimization reducing memory complexity by ~40× and computational complexity by ~10×.

KEY INSIGHTS

Combining semantic similarity with graph-based course authority provides a richer recommendation strategy while Top-K optimization significantly reduces the computational and memory requirements.

Yousef Tarek Nathan

AI & Data Science Developer