🎵 A Python-based content recommendation system utilizing ML algorithms to analyze a 600k-song dataset. It incorporates Direct Similarity, SVD, NMF, and Factorization Machines to deliver personalized music recommendations. The system handles cold start problems, optimizes using weighted similarity metrics, and includes tools for model visualization and performance evaluation.
- Multi-Model Engine: Developed a recommendation system that processes and analyzes a dataset of 600k songs, combining audio features and metadata. Utilized Direct Similarity Matching, SVD, NMF, and Factorization Machines to generate personalized recommendations based on user preferences and behavior.
- Optimization & Evaluation: Tackled the cold start problem, engineered cross-feature interactions, and implemented dimension-reduction and matrix factorization techniques. Created weighted similarity metrics that combine genre and audio feature analysis. Developed visualization tools to assess model performance, analyze latent spaces, and examine the effects of parameter changes on recommendations.