Researchers at the University of Claremont (United States) have developed an algorithm trained with neurological data from users and that is capable of predicting whether a song will be successful or not.
The experts from this institution have come to predict with a 97 percent accuracy rate whether a topic will be liked by the vast majority of users, using the listenersā neurophysiological responses according to different analysis models.
In this analysis, the complete work of which is published on the Frontiers scientific dissemination site, it is recalled that music is a format that directly influences the emotional states of people and that there are many aspects that achieve it, such as melody, tempo, key, or rhythm.
Emotional responses also come from multiple brain regions, as some are associated with emotion processing and others with long-term memory retrieval. The researchers, specifically, took as a reference the so-called peripheral networks, focused on feelings.
how was the experiment
For the sample, 33 participants between the ages of 18 and 57 were chosen, who listened to a total of 24 recent songs -13 of them considered hits and with more than 700,000 listeners in āstreamingā and 11 of them failures- and who were asked about their tastes and impressions of each of them.
Researchers at Claremont University set out to demonstrate that neurophysiological measures accurately identify hit songs, while usersā self-reported liking is not predictive.
Combined neurophysiology with automatic learning (āmachine learningā), an algorithm is created that āsubstantiallyā improves the classification of successful songs compared to traditional and linear statistical models.
After comparing the data, a new neuroprognostic approach was applied, which uses āmachine learningā applied to neural responses and has served to predict musical hits with great precision.
What would an algorithm for predicting musical hits be used for?
This approach, specifically, captures the neural activity of a small group of people to predict the effects of these songs en masse without having to measure the brain activity of hundreds of people.
In addition, the researchers concluded that neurophysiological responses within the first minute of the songs predicted hits with a success rate of 82 percent. This indicates that the first part of a song largely determines its popularity.
In their conclusions, the researchers comment that their intent with this study is āto show that neuroscience measures of the peripheral nervous system fairly accurately classify hits and missesā and that this approach can assess content value automatically.
āIf our findings are replicated, the ability to curate music and other forms of entertainment to give people what they want will enhance existing recommendation engines that will benefit artists, distributors, and consumers,ā they conclude.