🎧 Spotify Recommender System

A modular music recommender system that combines Content-Based Filtering and Collaborative Filtering to provide personalized track recommendations. The system is built using Python, DVC for pipeline management, and Streamlit for deployment. It is designed for scalability and reproducibility.

---

📌 Features

• ✅ Data Cleaning & Preprocessing
• 🎯 Content-Based Filtering using song metadata
• 👥 Collaborative Filtering using user listening history
• 🔀 Hybrid Recommendations
• 🧪 Modular and reproducible pipeline using DVC
• 🧱 Easy deployment with Streamlit and Docker
• 🧠 Uses encoders like TFIDF, OneHotEncoder, and MinMaxScaler

---

🗂️ Project Structure

spotify-recommender/
├── data/ # Raw and processed datasets
├── dvc.yaml # DVC pipeline stages
├── app.py # Streamlit frontend
├── content_based_filtering.py
├── collaborative_filtering.py
├── data_cleaning.py
├── hybrid_filtering.py
├── utils/ # Helper functions
│ └── encoders.py
├── models/ # Saved transformers/models
│ └── transformer.joblib
├── outputs/ # Final transformed data and matrices
├── requirements.txt
└── README.md

---

⚙️ Pipeline Overview

This project is managed using DVC to ensure a reproducible and modular ML pipeline.

🔹 Stages in dvc.yaml:

1. Data Cleaning

   • Input: Raw Spotify datasets
   • Output: Cleaned CSVs and fitted encoders

2. Content-Based Filtering

   • Input: Cleaned features
   • Output: Transformed feature matrices (.npz), similarity data

3. Collaborative Filtering

   • Input: User listening history
   • Output: Sparse interaction matrix, song-user similarity matrix

4. Hybrid Filtering

   • Input: Outputs from both previous stages
   • Output: Final ranked recommendations

---

🧠 Recommender Logic

📌 Content-Based: Vectorizes song metadata (genres, artists, lyrics)

Computes cosine similarity for ranking

📌 Collaborative: Uses user play history to form a song-user matrix

Calculates item-item similarity via cosine distance

🔄 Hybrid: Averages or weights both scores for better accuracy

---

🛠️ Setup & Installation

git clone https://github.com/VasuGx/Hybrid-Recommendation-System
cd spotify-recommender
pip install -r requirements.txt
dvc pull  # Pull data and models

---

📊 Sample Data

Song Metadata (songs.csv) User Listening History (user_history.csv)

---

#🧩 Tech Stack

Python DVC Scikit-Learn Dask Streamlit Docker GitHub Actions Amazon ECR

---

#✍️ Author

Vasu Gupta Designed with modularity and reproducibility in mind. @vasugupta1734@gmail.com