Accurately predicting hit songs using neurophysiology and machine learning

Area of Science:

• Neuroscience
• Music Psychology
• Machine Learning

Background:

• Identifying hit songs is a significant challenge in the music industry.
• Traditional methods focus on lyrical analysis of song databases.
• Previous approaches have limited predictive accuracy.

Purpose of the Study:

• To investigate the predictive power of neurophysiologic responses for identifying hit songs.
• To compare statistical and machine learning models for classifying music hits.
• To determine if the brain can rapidly identify hit music.

Main Methods:

• Measured neurophysiologic responses to a set of songs identified as hits and flops by a streaming service.
• Compared linear statistical models and ensemble machine learning approaches.
• Applied machine learning to neural data from the first minute of songs.

Main Results:

• A linear model using two neural measures achieved 69% accuracy in identifying hits.
• Ensemble machine learning on synthetic neural data achieved 97% accuracy.
• Machine learning on the first minute of neural response classified hits with 82% accuracy.

Conclusions:

• Neurophysiologic responses, when analyzed with machine learning, significantly improve the prediction of hit songs.
• The brain demonstrates a rapid ability to identify hit music within the first minute.
• Machine learning applied to neural data offers a powerful new method for predicting market success.