Who can know which song will become a hit? This algo can, and in an ultra-precise way!

Researchers have succeeded in developing an algorithm capable of predicting musical success. The method employs neurophysiological data collection and analysis using machine learning.

In the age of music streaming, tens of thousands of music titles are released worldwide every day. This massive influx of published music undeniably complicates the task of curating tracks for professionals in the sector. Whether this act of selection is part of online music listening platforms or the broadcasting schedule of a radio station, its goal remains the same: to unearth the songs that are going to be a hit!

97% Success in Tube Prediction

Considering that professionals in the field face a torrent of musical releases every day, researchers have seriously looked into the problem. Using machine learning technology combined with neurophysiological data, a research team in the United States has succeeded in developing a tool capable of predicting the success of music. For this, the 33 participants in the study were equipped with sensors and listened to a playlist of 24 titles. Their preferences as well as demographic data were also collected. The signals collected by the sensors reflect brain activity in relation to the mood and energy levels of the listeners.

By analyzing the data and using various methods of statistical approach, the researchers were able to make direct comparisons with the neurophysiological metrics of the experiment. One of the models used managed to predict a success rate with an accuracy of 69%. After applying a machine learning model, the hit prediction rate was 97% for listening to the song in its entirety, compared to 82% for the first minute of the song.

Streaming platforms could one day predict the success of a series

Paul Zak, a professor at Claremont Graduate University and lead author of the study, says such technology could make sorting music releases easier for professionals. Being able to identify the songs that are going to be a hit represents in a way the Holy Grail with streaming platforms, which can therefore be assured of satisfying their users. Rather than grapple with hundreds of potential picks daily, a computer algorithm could shortlist two or three titles. This would offer the professional a simplified, relevant and obviously faster choice.

According to Professor Zak, this methodology could probably be used far beyond music. The approach should find success and interest with video streaming services and the audiovisual entertainment industry in its broadest sense. Will movies and TV shows ever be released based on their potential to appeal? According to Zak, his approach is just as applicable in this field as in that of music lovers.

Source : TechXplore