PYFM - Henry Boisdequin

Music recommendation system built in Python, using the last.fm api.

Planning

Algorithm: Content-Based Filtering

1. 2 options for input:

• User information (username | uses favorite artists, genres, tags to offer genre/artist/track recommendations)
• Artist information (artist name | uses given artist to offer similar artist recommendations)

2. Calculate Item Similarity:

• Use Jaccard similarity & Cosine similarity to compute similarity between artist/genres/tracks

3. Build User Profile:

• Get weighted preferences for different user artist/genres/tracks based on last.fm metadata

4. Generate Recommendations:

• Use the similarity between an artist profile to other artist profiles OR user data to artist/genre/track profiles
• Rank the recommendations based on their similarity

Notes

API Key