Discovering the musical hits with Artificial Intelligence: 97% accuracy


The music hit prediction it is an arduous task. Popular streaming services like Spotify and Pandora they introduce new music to their users through methods like custom playlists. However, tracking the likelihood of which songs people will add to their playlists is challenging. A team of researchers from the United States has succeeded in developing a technique of machine learninga field of Artificial Intelligence, which can predict whether a song will be a hit with a staggering 97% accuracy.

Neural data-based approach

Various methods have been tried in the past to predict musical hits. From song lyric analysis to blog posts and social media mentions, various strategies have been explored. However, the predictive accuracy of these studies has been quite low. This research team decided to take a different approach, using the machine learning in brain responses of individuals when listening to music.

The brain responses of 33 individuals, ranging in age from 18 to 57, to 24 recent songs were studied. To record these responses, Rhythm + PPG heart sensors were used and participants were asked to share their preferences for each song.

Neuroprognostic technology

The researchers used a platform to measure neurophysiological responses that combine cues associated with attention and emotional resonance. This approach, known as “brain as predictor” or neuroprognosisallows capturing the neural activity of a small group of participants to predict outcomes at the population level.

Using a traditional linear logistic regression model, the researchers were able to predict with 69% accuracy which songs would be hits or flops. However, by applying a machine learning model to the neural data, this predictive accuracy increased to an astonishing 97%.

Implications for the music industry

This leap in predictive accuracy could have significant implications for music streaming services. Spotify, for example, has launched a similar AI-powered service that scans new releases to suit users’ tastes.

However, the increasing integration of Artificial Intelligence in music has not been welcomed with open arms by everyone. Some music industry players have raised concerns about the use of copyrighted music to train these machines, noting the need to obtain the necessary permissions.

More information at Frontiers.

Previous articleUnreal Engine 5.2 releases Electric Dreams demo for download
Next articleA company called Atari is releasing a brand-new 2600 cartridge this year
Brian Adam
Professional Blogger, V logger, traveler and explorer of new horizons.