OPTIMIZED SYSTEMS

Algorithm Predicts Music's Popularity

Rebecca Kutzer-Rice | Date: 12-29-11 | Comments

The top 40 song charts might seem like a jumble of notes and trends, but a new tool developed by scientists in the U.K. demonstrates that the popularity of music is less arbitrary than you might think. The scientists' algorithm can predict what songs will make it and what songs will be flops.

Today's computers are being used to predict everything from traffic to election results. New technology can now even predict what songs will be hits.

Researchers at the University of Bristol's Intelligent Systems Laboratory in the Faculty of Engineering analyzed the British top 40 singles chart from the past 50 years. The researchers sought to teach their computer to differentiate most popular songs (in the top five spots) from songs lower on the charts.

The algorithm can predict whether songs by stars like Lady Gaga will peak or plummet in the music charts.

They analyzed musical features, such as tempo, time signature, song duration and loudness, of the most popular and less popular songs. They also examined more complex song features, like harmonic simplicity, chord sequences and noise.

Using the 23 most important features, researchers then created a "hit potential equation," which scores a song based on how well its musical features align with past hits. Using the algorithm, the researchers have been able to categorize songs into "hit" and "no hit" with 60 percent accuracy.

"Musical tastes evolve, which means our 'hit potential equation' needs to evolve as well," Dr. Tijl De Bie, senior lecturer in artificial intelligence at Bristol, told PhysOrg.com. "Indeed, we have found the hit potential of a song depends on the era. This may be due to the varying dominant music style, culture and environment."

In addition to providing predictions about future songs, the research reveals some interesting facts about past decades. In the eighties, for instance, ballads and other slower styles were more likely to become hits than today. The researchers also found that popular songs are getting louder each year.

With the success of social media predictive tools, we at Smarter Tech wonder if Web 2.0 could make the research even more accurate. Many tweets about a certain band, for example, might indicate its popularity.

The study, which was presented at an international workshop in December, is the first research able to demonstrate an accurate "hit potential equation."

Check out the team's Website, Score a Hit, where you can try out the technology for yourself and see what songs have been most popular of the years.