Album Recommendation System Based on Random Forest Method and Mixed Effect Model

• Activity Number: 41
• Type: Contributed
• Date/Time: Sunday, August 9, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Taikgun Song* and Sanghoon Cho and Hyeongseon Jeon
• Companies: and Iowa State University and Iowa State University
• Keywords: Album Recommendation ; Mixed Effect Model ; Random Forest Method

Abstract

There are popular music website that recommends music such as Pandora or Last.fm. These popular music recommendation website uses two recommendation system commonly used in practice: collaborative filtering method and content based filtering method. In this study, we would like to improve the content based filtering method by introducing a new variable produced by random mixed effect model and lyric score. The new variable reflects the effect of the artist that cannot be explained by the specific variables, including genre, member composition, and internal information of the album. The lyrics score measures positivity of the words in lyrics. Then, the random forest method will be applied to estimate the review rate of the albums by using the training data that is based on the billboard chart from 2000 to 2013. Finally, an album could be recommended by probability which is driven from normalizing the estimated score.