Every day tens of thousands of new songs are launched on the market. That is why, given the enormous number of musical options, predicting which songs will make it to the playlists can be a difficult task for radio stations and different streaming services. streamingas Spotify and Apple Music.
However, as other industries have already done, music companies have been making use of artificial intelligence (AI). However, until recently, the accuracy that had been obtained with these technological tools had been barely 50%, that is, half the chance that a song would succeed, and half that it would go unnoticed.
However, a team of researchers from USAusing an AI, managed to identify with 97% accuracy the songs that will be a success on radio stations and different streaming platforms.
“By applying machine learning to neurophysiological data, we were able to almost perfectly identify hit songs,” he said. Paul Zakprofessor at Claremont Graduate University and lead author of the study published in Frontiers in Artificial Intelligence.
To carry out the study, the 33 participants were fitted with sensors and 24 songs were played. Once that was done, the scientists asked them for their preferences and for certain demographic data. It was thus that they measured the neurophysiological responses of the volunteers to the songs that had been played for them.
“The brain signals we collected reflect the activity of a brain network associated with mood and energy levels,” Zak said.
This made it possible for experts predict market outcomes, including the number of streams for a songThis is based on the data of a few.
The method used by storytellers to predict musical hits is known as “neuroprevision”, which consists of capturing the neural activity of a small group of people in order to predict the effects at the population level without having to measure the brain activity of all subjects.
After collecting the data, the researchers used different statistical approaches in order to assess the predictive accuracy of the neurophysiological variables, which allowed them to directly compare the models. In addition to this, to improve predictive accuracy, trained an ML model which tried different algorithms to arrive at the best results.
It was in this way that, by applying machine learning to the collected data, they realized that the success of predicting the songs that would be a hit rose to 97%.
“The neural activity of 33 people being able to predict whether millions of others listened to new songs is quite astonishing. Nothing close to this accuracy has ever been shown,” the professor remarked.
#Artificial #intelligence #predicts #successful #songs #accuracy