Recommending Collaborations for New and Established Artists

Frequent Itemset Analysis and Recommender Systems

Crisanto Chua Jeddahlyn Gacera

Abstract

We consider the problem of talent managers of new artists. The goal is to make their clients popular. One of the most important initial activities of any competent manager is to properly launch the careers of their clients. This could be achieved by the new artists doing collaborations with established ones to make them more popular to the public. The question this study tries to investigate is who should they partner with? This work aims to solve this by exploring similarities between established and new up-and-coming artists based on user taste profiles provided by the Echo Nest Subset of the Million Song Dataset. Furthermore, applying Frequent Itemset Analysis and Recommender Systems to suggest possible established artists who they can front act for in concerts and shows. This study will focus on artists from the rock genre which is again becoming popular because of recent movies about rock stars such as Queen (Bohemian Rhapsody) and Elton John (Rocketman).

Introduction

We are all familiar with the term “struggling artist”. This describes a creative person who, because of his/her love for art, is not able or is struggling to make ends meet. According to a recent report titled “By Artists, For Artists”, nearly 50 percent of those artists surveyed in Massachusetts said they had “business losses from their creative practices (Massachusetts Artists Leaders Coalition, 2018). Even more, they need to work outside their skill and craft just to survive. Sometimes, all that these artists need is a “break” via exposure to the public. The best way to achieve this is for these new artists to perform in a concert or a show by an already established/veteran artist and/or be featured in their albums.

Online music platforms like Spotify and Pandora provide a wealth of data that could help in giving recommendations as to which artists to collaborate with. This recommender system could play an important role for music executives to identify possible partnerships between established artists and budding performers.

Methodology

The following steps were taken for this study:

  • Problem identification and review of related literature
  • Collection and description of music data from Million Song Dataset website
  • Data preparation and selection
  • Exploratory data analysis (EDA)
  • Performing Frequent Itemset Mining (FIM)
  • Building a Recommender System (RS)

If you wish to have a copy of the technical paper for this project, kindly contact us via e-mail or LinkedIn.