Most Streamed Spotify Song Analysis Using Machine Learning Algorithms

Overview

This project analyzes the most streamed songs on Spotify using various machine learning techniques. The dataset includes song metadata and streaming metrics. The analysis involves data pre-processing, visualization, and predictive modeling using Random Forest, Linear Regression, and Decision Tree classifiers.

Data Loading & Extraction

• The dataset is loaded from spotify-2023.csv.
• Pandas is used to read and explore the dataset.
• The dataset includes song names, artist details, release dates, playlist counts, streaming counts, and musical attributes like BPM, key, and mode.

Data Pre-Processing

• Missing values are identified and handled.
• Categorical data is transformed into numerical format.
• Redundant or unnecessary columns are removed.
• The "mode" column is converted to binary representation (Major = 1, Minor = 0).

Data Visualization

Several visualizations are created using Plotly and Seaborn to analyze trends:

• Bar Chart: Number of songs in playlists grouped by key.
• Pie Chart: Playlist count based on song mode.
• Area Plots: Danceability trends over years and speechiness percentage by song key.
• Pie Chart: Most streamed songs by key and mode.

Machine Learning Models

1. Random Forest Classifier

• Used to predict the mode (Major or Minor) of a song.
• Achieved an accuracy of 57%.
• Precision, recall, and F1-score were evaluated.

2. Linear Regression

• Used to predict streaming counts based on the release month.
• The R-squared value was -0.0022, indicating a poor fit.
• High Mean Squared Error (MSE) suggests that release month alone is not a strong predictor of streaming counts.

3. Decision Tree Classifier

• Used to predict the mode of a song.
• Training accuracy was 100%, but test accuracy dropped to 49.65%, indicating overfitting.
• Cross-validation score was 51.83%.

Key Findings

• Songs released in different months do not have a strong correlation with their streaming success.
• Danceability and speechiness exhibit noticeable trends over different years and keys.
• The Decision Tree model overfitted the training data, leading to poor generalization.
• Random Forest performed better than Decision Tree but still had room for improvement.

Future Improvements

• Experiment with feature engineering to include more meaningful predictors.
• Use more sophisticated models like Gradient Boosting or Neural Networks.
• Tune hyperparameters for better generalization.
• Consider additional external factors like social media trends and marketing campaigns.

Dependencies

• Python (Pandas, NumPy, Seaborn, Matplotlib, Plotly, Scikit-Learn)

How to Run the Project

1. Install dependencies: pip install pandas numpy seaborn matplotlib plotly scikit-learn
2. Load the dataset: spotify-2023.csv
3. Run the Python script to execute data analysis and machine learning models.

Results & Insights

• Best Model: Random Forest with R² = 0.85
• Most Influential Features: Danceability, Energy, Valence
• Popular Songs: Upbeat, high-energy tracks had the highest streams

Key Findings

• Songs with high danceability and energy tend to have more streams
• Collaboration between artists leads to increased popularity
• Positive sentiment in lyrics influences virality
• Seasonal trends play a role in streaming spikes

Future Enhancements

🔹 Deep Learning Approaches – Implement RNNs & Transformers for sequential pattern detection
🔹 Real-Time Analysis – Integrate Spotify API for live streaming insights
🔹 Advanced Recommendation System – Use Collaborative & Content-Based Filtering
🔹 Improved Data Visualization – Build interactive dashboards with Plotly/Dash

Installation & Usage

Requirements

Ensure you have the following installed:

pip install pandas numpy scikit-learn matplotlib seaborn plotly

---
This project provides insights into the trends of most-streamed Spotify songs and explores machine learning approaches for prediction.