Project Title: Personalized Music Recommendation System

Program: MIT Applied Data Science Program (ADSP) Capstone — qualified as E, culminating in an oral exposition to industry professionals with an accompanying PowerPoint presentation.

Overview:
This capstone focused on building a personalized music recommendation system that suggests relevant songs to individual users based on their listening behavior. The system predicts a user’s top-10 likely preferences by analyzing patterns in large music datasets.

Key Components:

Data Source & Preprocessing: Utilized large music interaction datasets (e.g., user play counts and song metadata) and performed cleaning, filtering, ID encoding, and rating scaling to prepare data for modeling.

Modeling Approaches: Explored multiple recommendation strategies, including popularity baselines, collaborative filtering (user–user, item–item, matrix factorization like SVD), cluster-based methods, and content-based techniques.

Evaluation: Assessed performance using standard metrics such as Precision@k, Recall@k, and F1-score@k to compare models’ effectiveness in recommending top-N songs.

Outcome:
The final system demonstrated strong recommendation quality, with optimized latent-factor models (e.g., SVD) providing personalized suggestions tailored to users’ historical preferences. Evaluation results helped justify the selected model as a practical, scalable solution.

Presentation & Defense:
Results were summarized in a PowerPoint deck highlighting objectives, data preprocessing, model comparisons, performance evaluation, and future work. The capstone concluded with an oral defense to industry professionals, explaining the methodology, results, and real-world applicability of the recommendation engine.