AI can now predict hit songs with 97% accuracy

Various methods like analyzing song lyrics, blog postings, social media mentions, and brain activity have been experimented with before to predict hit music. Yet, predictive accuracy for most studies is quite low, said the researchers in their study. Citing an example, the researchers quoted a previous study that used functional MRI to predict music popularity with predictive accuracy below 50%.

Mapping neural data via machine learning 

They applied the machine learning technique to the brain responses of 33 individuals within the 18-57 age range. The participants were fitted with Rhythm + PPG cardiac sensors and were made to listen to 24 recent songs. They were asked about their preferences for each one. The participants also took a survey on demographics.

The team then used a platform to measure the neurophysiologic responses. The platform combines signals associated with attention and emotional resonance. This is called  “brain as predictor” or “neuroforecasting,” an approach that captures neural activity from a small group of participants to predict population outcomes.

“Our analysis showed that two measures of neurophysiologic immersion in music identified hits and flops with 69% accuracy using a traditional linear logistic regression model,” said the researchers in the study.

However, applying a machine learning model to neural data improved the predictive accuracy from 69% to 97%.

The team concluded that their findings will pave the way for streaming services to build customized playlists more effectively and “to give people just what they want will improve existing recommendation engines benefiting artists, distributors, and consumers.”

AI seeping into the industry

Spotify launched a similar service named DJ. The AI-powered service scans new releases to match users' likes and dislikes and provides commentary about the upcoming song and artists, reported Interesting Engineering in February.

However, not everyone in the music industry is happy with the increasing integration of AI with music. Universal Music Group had sent a strongly worded email to Spotify, Apple Music, and others to not let AI companies access copyrighted music “without obtaining the required consents” to train their machines. 

Study abstract:

Identifying hit songs is notoriously difficult. Traditionally, song elements have been measured from large databases to identify the lyrical aspects of hits. We took a different methodological approach, measuring neurophysiologic responses to a set of songs provided by a streaming music service that identified hits and flops. We compared several statistical approaches to examine the predictive accuracy of each technique. A linear statistical model using two neural measures identified hits with 69% accuracy. Then, we created a synthetic set data and applied ensemble machine learning to capture inherent non-linearities in neural data. This model classified hit songs with 97% accuracy. Applying machine learning to the neural response to 1st min of songs accurately classified hits 82% of the time showing that the brain rapidly identifies hit music. Our results demonstrate that applying machine learning to neural data can substantially increase classification accuracy for difficult to predict market outcomes.