Abstract: In this work, we endeavor to take care of the Hit Song Science issue, which plans to foresee which melodies will become diagram besting hits. We develop a dataset with around 1.8 million hit and non-hit tunes and removed their sound elements utilizing the Spotify Web API. We test four models on our dataset. Our best model was arbitrary woods, which had the option to anticipate Billboard melody accomplishment with 88% exactness. In the current review, we moved toward the Hit Song Science issue, planning to anticipate which tunes will become Billboard Hot 100 hits. We grouped a dataset of roughly 4,000 hit and non-hit tunes and extricated every tunes sound highlights from the Spotify Web API. We had the option to anticipate the Billboard accomplishment of a tune with around 75% precision on the approval set, utilizing four AI calculations. The best calculations were Support vector machine, Logistic Regression and a Deep learning.
Keywords: Machine Learning ,Hit Song Science ,Classification ,Data Mining, Data Collection
| DOI: 10.17148/IJARCCE.2021.101213