In a world awash with music, it is becoming increasingly difficult for streaming services and radio stations to select which songs to add to their playlists. Despite efforts to combine human listeners and artificial intelligence, the accuracy rate currently sits at 50%, making it difficult to reliably predict hit songs.
Despite those indicators, researchers in the United States have used a machine learning technique based on brain responses and have managed to predict hit songs with an astonishing 97% accuracy.
The power of machine learning and neurophysiological data
In a study carried out by researchers at Claremont University, participants were fitted with special sensors to measure their brain responses while listening to a set of 24 songs.
The scientists focused on the participants’ neurophysiological responses, which reflect the activity of a brain network associated with mood and energy levels. This data made it possible to accurately predict market outcomes, including the number of plays a song might achieve.
The methodology that led to surprising results
The approach used in this study, known as “neuroprognostic,” captures the neural activity of a small group of people to predict effects at the population level, without having to measure the brain activity of hundreds of individuals.
The researchers used various statistical approaches and trained a machine learning model with different algorithms to get the most accurate results. They found that a linear statistical model was able to identify hit songs with a 69% success rate. However, by applying machine learning to the collected data, this rate was raised to 97%. Even when analyzing neural responses from the first minute of songs, the model achieved an 82% success rate in identifying hits.
Implications and future applications
The near perfect precision of this approach has important implications for the music industry and streaming services.
With the ability to easily identify new songs that are likely to become hits, streaming services can improve their efficiency when selecting songs for playlists, which in turn can further satisfy listeners.
Furthermore, the researchers suggest that this methodology could be applied beyond music, such as in predicting hits in movies and TV shows.
Challenges and limitations
Despite the promising results, it is important to mention that this study had some limitations. The analysis was based on a relatively small number of songs and the participants represented moderate diversity in terms of demographics, excluding certain age and ethnic groups.
However, the researchers highlight the methodology as the main contribution of their study and suggest that this approach can be used in other fields of entertainment.
What do music and the brain have in common? The key to predicting hit songs
Using machine learning based on brain responses opens up new possibilities for predicting hit music and potentially other forms of entertainment.
The 97% precision obtained in this study demonstrates the potential of this technique. If wearable neuroscience technologies become more common in the future, we could see a more personalized approach to the delivery of entertainment content, where more limited options based on individual neurophysiology would be offered.
Ultimately, this could improve the consumer experience and simplify decision-making in a world increasingly saturated with choice.
