🎵 Spotify Listening Data Analysis & BERT-Based Song Recommendation

This project analyzes personal Spotify listening history and uses a BERT-based model to recommend similar tracks based on your listening behavior and track metadata.

---

📌 Features

• 📊 Exploratory Data Analysis:

  • Most listened artists and albums
  • Listening patterns by time
  • Track skip rate
  • User behavior insights

• 📈 Visualizations:

  • Artist frequency bar plots
  • Listening time histograms
  • Hourly activity line plots
  • Pie charts of listening reasons

• 🧠 BERT-Based Recommendation System:

  • Uses BERT embeddings to find similar tracks
  • Generates song recommendations based on cosine similarity
  • Engagement label prediction (Low, Medium, High)

---

🗂️ Dataset

• File Used: spotify_history.csv which is in kaggle

• Contains:

  • track_name, artist_name, album_name
  • ms_played, ts (timestamp), reason_start, reason_end, skipped

---

🛠️ Tech Stack

• Python 🦀
• pandas, numpy
• matplotlib, seaborn
• PyTorch, HuggingFace Transformers 🤗
• scikit-learn

---

🚀 How to Run

1. Clone the repository:

   git clone https://github.com/Raghuraam21/SongRecommandation.git

🧠 How It Works

1. Preprocessing:

   • Extracts features like hour, minute, and total play time
   • Generates a text field from track, artist, album
   • Labels engagement levels (Low < 10s, Medium, High > 60s)

2. BERT Embedding:

   • Uses bert-base-uncased to generate vector embeddings from text
   • Compares track embeddings using cosine similarity

3. Recommendation:

   • Given a seed song, finds top-k most similar tracks

---

📌 Example Output

Based on your listening to: 'Born To Die' by Lana Del Rey  
Recommended tracks:  
1. Summertime Sadness by Lana Del Rey (similarity: 0.943)  
2. Young and Beautiful by Lana Del Rey (similarity: 0.938)  
3. Video Games by Lana Del Rey (similarity: 0.927)

📬 Contact

Made by Raghuraam