EDA On Attribute variation with spotify sub genres
Different Approaches:
- Content Based: Song are recommended based on content similarity to other songs
- Collaborative: Models recommends based on preferences of similar users
Data Preprocessing
Find relevant columns and how to weigh them
- Numeric
| track_popularity | year | danceability | energy | key | loudness | mode | speechiness | acousticness | instrumentalness | liveness | valence | tempo | duration_ms |
|---|
- String / Categorical
| track_artist | playlist_genre | playlist_subgenre |
|---|
Normalize / Standardize the numeric data
Vectorize the categorical data
Data Processing
Perform PCA
Save each song as its combination of PCA
Finding Similarity
Perform cosine similarity operations and find most similar songs