AI predicts musical hits with near-perfect accuracy
It can be challenging for streaming services and radio stations to select which songs to add to their playlists, given the sheer number of new releases every day. To determine which songs will appeal to a broad audience, these services have used a combination of human listeners and artificial intelligence. However, this method has only achieved a 50% accuracy rate in predicting hit songs. In the US, researchers have recently developed a comprehensive machine-learning approach that analyzes brain responses, resulting in a 97% accuracy rate in predicting musical hits.
“By applying machine learning to neurophysiologic data, we could almost perfectly identify hit songs,” said Paul Zak, a professor at Claremont Graduate University and senior author of the study published in Frontiers in Artificial Intelligence. “That the neural activity of 33 people can predict if millions of others listened to new songs is quite amazing. Nothing close to this accuracy has ever been shown before.”
How the study was conducted
As part of a study, participants were given off-the-shelf sensors and asked to listen to a set of 24 songs. The researchers then collected data on the participants’ preferences and demographics while simultaneously measuring their neurophysiological responses to the music. The results showed that the collected brain signals reflected activity in a brain network associated with mood and energy levels, which allowed the researchers to use a technique called ‘neuroforecasting’ to predict market outcomes, such as the number of song streams, based on data from just a few participants.
To assess the predictive accuracy of the neurophysiological variables, the researchers conducted different statistical analyses and trained a machine learning model using various algorithms to arrive at the highest prediction outcomes. They found that a linear statistical model was able to identify hit songs with a success rate of 69%. However, when machine learning was applied to the data collected, the rate of correctly identified hit songs increased to an impressive 97%.
Additionally, the researchers used machine learning to analyze the neural responses to the first minute of each song, which resulted in a successful hit rate of 82%. Overall, these findings demonstrate the potential for using neuroforecasting and machine learning to predict market outcomes with minimal input data accurately.
Paul Zak, a professor at Claremont Graduate University, explained,
“This means that streaming services can readily identify new songs that are likely to be hits for people’s playlists more efficiently, making the streaming services’ jobs easier and delighting listeners.”
Despite the near-perfect prediction results of his team, the researchers pointed to some limitations. For example, they used relatively few songs in their analysis. Furthermore, the demographics of the study participants were moderately diverse but did not include members of certain ethnic and age groups.
“Our key contribution is the methodology. It is likely that this approach can be used to predict hits for many other kinds of entertainment too, including movies and TV shows.”
In-Article Image CreditsMan dressed like robot listening to music through headphones via Dream Studio with usage type - Self
Robot woman playing a musical piece on a piano via Dream Studio with usage type - Self
Artificial Intelligence Man in suit playing music on a keyboard via Dream Studio with usage type - Self
Featured Image CreditArtificial Intelligence Man in suit playing music on a keyboard via Dream Studio with usage type - Self