Year after year, a staggering 90% of new single releases never make it onto the charts, with only a minute portion achieving true hit status. The music industry largely remains in the dark as to why this is, except for a scientific team in California who claim they can identify potential hit songs with a 97% success rate.

Every single day, the world is flooded with tens of thousands of fresh tracks. This relentless flood of choices presents a challenging situation for streaming platforms and radio stations when deciding which tracks should feature on their playlists. They’ve traditionally relied on human listeners and AI to identify those tracks likely to strike a chord with a broad listener base. However, this strategy only yields about a 50% success rate, providing an unreliable method of predicting which songs will emerge as hits.

A breakthrough has been made by a group of US-based researchers who have applied an advanced machine-learning technique to the analysis of brain reactions. With this approach, they’ve succeeded in predicting potential hit songs with an impressive accuracy rate of 97%.

“By applying machine learning to neurophysiologic data,” comments senior author Prof. Paul Zak, “we could almost perfectly identify hit songs.

“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.”

The findings of the study were published in Frontiers in Artificial Intelligence.

In the study, participants were fitted with readily available sensory devices, asked to listen to a group of 24 songs, and questioned about their preferences along with some personal demographic information. Throughout this process, the team recorded the participants’ neurophysiological reactions to the songs.

“The brain signals we’ve collected reflect activity of a brain network associated with mood and energy levels,” explains Zak.

This information enabled the research team to predict market outcomes, like a song’s stream count, based on limited data.

The researchers term this method ‘neuroforecasting.’ It involves monitoring neural activity from a small group to predict population-level behavior, negating the need to measure brain activity from a multitude of individuals.

Once the data collection was complete, the researchers employed varying statistical methods to evaluate the predictive precision of the neurophysiological variables. This made it possible to directly compare the different models. To enhance predictive accuracy, they developed a machine learning (ML) model that experimented with various algorithms to attain the highest prediction outcomes.

They discovered that a linear statistical model could pinpoint hit songs with a 69% success rate. However, the application of machine learning to the collected data boosted the correct hit song identification rate to 97%. Moreover, when they applied machine learning to neural responses during the first minute of the songs, the hit song identification success rate was 82%.

“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,” adds Zak.

“If in the future wearable neuroscience technologies, like the ones we used for this study, become commonplace, the right entertainment could be sent to audiences based on their neurophysiology,” points out the author.

“Instead of being offered hundreds of choices, they might be given just two or three, making it easier and faster for them to choose music that they will enjoy.”

While the research team produced almost flawless prediction results, they acknowledged a few limitations. For instance, their analysis included a relatively limited number of songs. Additionally, the participant pool, although moderately diverse, did not encompass representatives from certain ethnic and age groups.

However, they anticipate that their methodology has potential applications beyond merely identifying hit songs, in part because of its ease of implementation.

“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,” concludes Zak.

Image Credit: PG/Bauer-Griffin/GC Images Via Getty Images