Can you Predict a Hit? Finding potential hit songs through last.FM predictors

Abstract

In the field of Music Information Retrieval, studies on Hit Song Science are becoming more and more popular. The concept of predicting whether a song has the potential to become a hit song is an interesting challenge. This paper describes my research on how hit song predictors can be utilized to predict hit songs. Three subject groups of predictors were gathered from the Last.fm service, based on a selection of past hit songs from the record charts. By gathering new data from these predictors and using a term frequency-inversed document frequency algorithm on their listened tracks, I was able to determine which songs had the potential to become a hit song. The results showed that the predictors have a better sense on listening to potential hit songs in comparison to their corresponding control groups.