Predicting and Composing a Top Ten Billboard Hot 100 Single with Descriptive Analytics and Classification

In late 20th and early 21st century Western popular music, there are cyclical structures, sounds, and themes that come and go with historical trends. Not only do the production techniques utilized reflect technological advancements (the Yamaha DX7, the Roland 808, etc.), the art form reflects contemporary cultural attitudes through lyrics and stylistic choice. Through this lens, pop songs can serve as historical artifacts for their unique ability to captivate listeners based on their generally acceptable and familiar elements, both upon release and with future audiences. It raises the questions: “Can a chronological analysis of artistic choices reveal trends in songwriting and popular music composition?”; “Based on collected analysis, could forecast data suggest criteria that a future hit song may fit?”; and “How could the next ‘hit song’ sound, based on the calculated criteria from trend analysis and forecasting techniques?” By manually listening to and analyzing Billboard songs for each of the last 50 years and employing an assortment of feature selection and classification techniques, a random forest model predicts some of the significant characteristics of a potential future hit song. This prediction provided the framework for an original composition