Hybrid Music Recommendation System combining Neo4j GraphDB based and ML-based Approach

Due to the significant expansion of the internet, particularly the World Wide Web, there is now an extensive collection of music accessible online. Recognizing the vastness of this musical landscape, it has become imperative to go beyond traditional search functionalities and introduce a service that not only facilitates user searches but also provides tailored music recommendations. This research paper presents an innovative approach to music recommendation by combining machine learning (ML) techniques with graph database technology. The proposed system utilizes ML to generate recommendations based on audio features extracted from a vast music dataset. The recommendation engine employs a clustering pipeline and cosine distance metrics to identify songs with similar characteristics. Additionally, the system leverages the capabilities of Neo4j, a graph database, to store and query relationships between songs deemed similar by their audio features. The integration of Neo4j enhances the recommendation process by allowing the system to consider not only audio feature similarities but also graph-based relationships. This hybrid approach aims to provide users with more diverse and contextually relevant music recommendations. The paper outlines the architecture of the system, including data preprocessing, ML model training, and the incorporation of Neo4j for storing and querying music relationships. Evaluation results demonstrate the effectiveness of the combined ML and graph-based recommendation system, offering a promising direction for advancing personalized music recommendation systems. This research contributes to the intersection of ML and graph databases in the field of music recommendation, offering a novel perspective on enhancing recommendation quality through diverse data representations.