Classification of music into moods using musical features

Usually, music is generally classified on the basis of its genre which indicates its musical style or musical form based on some sort of shared history. On the contrary, this paper aims to classify a given track into a mood such as happy, sad, peaceful and angry rather than based on its genre because more often than not, the listener prefers to hear songs similar to each other both in terms of the mood it evokes in the listener and otherwise. This is done by using the valence and arousal values that came along with the 1000 song database that we are using, as well as extraction of several musical features such as tempo, energy, pitch, etc. Using a combination of all these features, we use 4 different classification algorithms namely - Support Vector Machine (SVM), Naive Bayes, Linear Discriminant Analysis and Decision Trees to classify the music into the aforementioned moods and the accuracy achieved with these classifiers is 59.8%, 75.4%,78.5% and 71.4% respectively.