Spotify Popularity Classification using KNN

This project classifies Spotify tracks as Popular or Unpopular using the K-Nearest Neighbors (KNN) algorithm. It demonstrates data preprocessing, model training, evaluation, and interpretation using real-world audio feature data.

Objective

Build a binary classification model to predict song popularity based on various audio features, helping to understand what makes a song popular.

Dataset

• Source: [Spotify Tracks Dataset by Maharshi Pandya]
• (www.kaggle.com/datasets/maharshipandya/-spotify-tracks-dataset)
• Size: Over 114,000 tracks
• Target Variable: popularity (binarized as Popular ≥ 70, else Unpopular)
• Features Used:
  • danceability
  • energy
  • loudness
  • speechiness
  • acousticness
  • instrumentalness
  • liveness
  • valence
  • tempo
  • duration_ms

Workflow Summary

1. Data Preprocessing

   • Handled missing values
   • Converted popularity into binary target variable
   • Normalized numerical features using MinMaxScaler

2. Model Implementation

   • Used KNeighborsClassifier from scikit-learn
   • Tried different k values (from 3 to 9)
   • Split data using train-test split

3. Evaluation

   • Measured performance using accuracy score
   • Analyzed confusion matrix
   • Identified top-performing value of k
   • Performed feature importance analysis via permutation method

4. Visualization

   • Confusion matrix heatmap
   • Accuracy vs K plot

Results

• Best Accuracy: Achieved approximately 96% accuracy at k = 9.
• Key Observations:
  • Features like danceability, energy, and valence had high influence.
  • Model performance was sensitive to normalization.
  • KNN showed decent results but would benefit from more advanced techniques or balanced class distributions.

Technologies Used

• Python
• Pandas
• Scikit-learn
• Matplotlib
• Seaborn

Potential Enhancements

• Use multi-class genre classification instead of binary popularity.
• Integrate a Streamlit app for interactive predictions.
• Try other ML algorithms (Random Forest, Logistic Regression) for comparison.
• Handle class imbalance more effectively.