WEBSITE REKOMENDASI DAN KLASIFIKASI LAGU MENGGUNAKAN METODE WEIGHTED K-NEAREST NEIGHBOR

Abstract

As the years went by, music has become one of the most evolving aspects of human history. There is a load of musical development around the globe, especially in music genres. Due to these differences and developments, a design was created to be able to make song recommendations according to the genre types and classifications of music or song. The data that is processed as training data is in the form of song metadata with various music features sourced from Spotify. Song recommendations are performed using the Euclidean Distance calculation between musical features or songs, while song classification is carried out using the Weighted K-Nearest Neighbor (WKNN) method calculation through audio wave type file analysis which then takes the musical features and calculates them based on the existing song or music data. The end result of this process is the genre class label. There is also a classification evaluation calculation using a confusion matrix. With the design of this system, it is hoped that the user will be able to search for song recommendations that have similarities to the song chosen by the user and classify genres according to the user's input song.