Mining Music Artist Similarity based on MetaPath

Mar 2015 - May 2015  |  

Course instructor: Jiawei Han

Team: Mingrui Chen

Fixed Dimensions

Images with fixed dimensions
  1. Tool Interface

  2. Two Meta-Paths

  3. Heterogeneous Networks

  4. Database Sketch for Meta-Path ATPTA

This is a course project in CS512 Data Mining Principles at UIUC. Some widely studied methods for music artist similarity measure include collaborative filtering and using co-occurrence of music/artists in reviews or playlist. In this course project, we tried to involve meta-paths and apply PathSim into artist similarity measure meta paths are involved into artist similarity measure. We built heterogeneous networks and measured artist similarity based on two meta paths: a playlist-based one Artist-Track-Playlist-Track-Artist (ATPTA) and a track-based one Artist-Track-Genre-Track-Artist (ATGTA). A visualization interface was built to display the similarity search results.