(ML) Music Recommendation System

🎥 Model Diagram

📖 Presentation Material

https://docs.google.com/presentation/d/1LFrR0mcRD333V0jMQCteKuijk77TETDP/edit?usp=drive_link&ouid=113654655373866451129&rtpof=true&sd=true

A recommendation program for "music". The user enters two songs, and the main class 'Music_Recommender' recommends a list of songs relative to the entered two songs.

Three filtering models were integrated to develop the music recommendation system.

• Collaborative Filtering – KNN Classifier
• Content Based Filtering – Lyric based Cosine Similarity
• Content Based Filtering – Music Features based Sigmoid calculation

Dataset:

• https://www.kaggle.com/datasets/maharshipandya/-spotify-tracks-dataset
• http://millionsongdataset.com/
• Real-time crawling from Melon to obtain the latest information

👬 Team Composition

Model Development (2), Presentation Material (2)

🔨 Role and Function

• Collaborative Filtering, Content Based Filtering, Content Based Filtering
• Configuring a big all-in-one function

⚒️ Used Technologies and Libraries

• Collaborative Filtering – KNN Classifier
• Content Based Filtering – Lyric based Cosine Similarity
• Content Based Filtering – Music Features based Sigmoid calculation
• Crawling - BeautifulSoup4

💡 Reflections

• Developing a recommendation system for my favorite field, music, by directly obtaining scattered song data from the internet was a novel experience.
• However, it was disappointing to find that there is not as much music data as expected.
• It was unfortunate that the tf-idf library supports only English.
• Establishing criteria for providing optimal recommendations for two songs was challenging.
• Integrating three models made it difficult to decide on the basis for assigning the order of results from the models.