Music Analysis & Recommendation System (M.A.R.S)

By Kashish Kohli, Kanksha Masrani

Can machine learning algorithms predict how successful a song will be on the global stage? Can they create music phenomena like Rihanna and Michael Jackson? With the global music industry worth over $130 billion US dollars, it would certainly be interesting to know! Big data students Kashisk Kohli and Kanksha Masrani set out to answer this and other questions with their CMPT 733 final project: a music analysis and recommendation system (MARS). Using six different machine learning algorithms, MARS attempts to predict a song's popularity based on features like loudness and danceability, boasting an accuracy rate of 70%. Another feature of MARS is the global sentiment analyzer, a tool that sheds light on which countries prefer songs with a negative or positive undertone. This allows artists to predict the success of a song in a given country depending on the tone of its lyrics. Finally, MARS offers personalized song recommendations for music fans. It makes recommendations based on all-time favourite songs among all users as well as songs that users with similar tastes have liked. Kohli's and Masrani's project showcases sophisticated data science skills and is an impressive one-stop shop for music producers, artists, and consumers.

Project report | Project poster