Looks like the music industry isn't called a hit factory for nothing: it appears a computer algorithm could correctly predict which tunes were the top 10 blow-out pop hits in the year when fed a song for testing.

Developed by scientists at the University of Antwerp in Belgium, the algorithm is still in its initial stages. Lead by scientist Dorien Herreman, the research team developed the algorithm after research they conducted on dance hits spanning over the past 30 years, from 1985 up until the present day. When fed a dance song as a sample, the algorithm had a 65 percent success rate of properly picking which tune would be a hit (i.e., achieve a top 10 slot in the music charts), and which would be a flop, as well as a 70 percent chance of picking six bonafide successful pop songs out of 10.

Here are their stats for the Billboard 2015 bonafide hits:

Billboard 2015 Hot Dance/Electronic Songs

1. "Lean On" by Major Lazer & DJ Snake Featuring M0 - 82%
2. "Where Are U Now" by Skrillex & Diplo With Justin Bieber - 72%
3. "Hey Mama" by David Guetta Featuring Nicki Minaj, Bebe Rexha & Afrojack -72%
4. "You Know You Like It" by DJ Snake & AlunaGeorge - 63%
5. "Waves" by Mr. Probz - 68%
6. "Outside" by Calvin Harris Featuring Ellie Goulding - 82%
7. "Prayer In C" by Lillywood & Robin Schulz - 65%
8. "Blame" by Calvin Harris Featuring John Newman - 88%
9. "How Deep Is Your Love" by Calvin Harris & Disciples - 62%
10. "I Want You To Know" by Zedd Featuring Selena Gomez - 89%

The algorithm also tested out the official top 10 singles of the year on the U.K. charts:

Official Charts Company 2015 Singles

1. "Happy" by Pharrell Williams - 83%
2. "Rather Be" by Clean Bandit - 74%
3. "All Of Me" by John Legend (not dance)
4. "Waves" by Mr Probz - 68%
5. "Thinking Out Loud" by Ed Sheeran (not dance)
6. "Ghost" by Ella Henderson - 79%
7. "Timber" by Pitbull ft. Kesha - 90%
8. "Stay With Me" by Sam Smith (not dance)
9. "Let It Go" by Idina Menzel (not dance)
10. "All About That Bass" - Meghan Trainor - 87%

To work, the algorithm analyzes particular components that make up the song in question, which the team derived from Echo Nest, a music intelligence software that breaks down these audio characeristics for search engines in apps like Spotify and Apple Music. The team then parsed 139 different characteristics (aspects like song length, tempo, bass, and so forth), and "even more intangible qualities like a song's timbre or tone color."

"Some other research tried to build models on pop songs, yet with bad results," said Herremans in an interview with Motherboard. "Dance charts seemed to be songs in the most similar style, versus pop charts that can contain both rock or dance songs. By choosing one particular style, we can gather more directed features and get better prediction."  

"The hit predictions seem very good! Even better than the ones we did in the paper," Herremans added. The research paper, "Dance Hit Song Prediction," was published in 2014 in the Journal of New Music Research.

Via: Motherboard

Photo: The Republic of Korea | Flickr