Data analysis of music preferences of web users based on social and demographic factors

Abstract

This study deals with the present-day problem of analyzing data of music preferences of web users based on social and demographic factors. The paper also analyzes the basic parameters of the user profile, which include such parameters as music preferences and movie preferences, personality traits, hobbies and interests, views on life, spending habits, phobias and health habits, and opinions, and demographics. The missing data imputation is carried out by multivariate imputation by chained equations. For analysis, 257 attributes are taken into account. Classification and Regression Trees were used as an algorithm for tree construction to find related variables to target variables "Opera" and "Classical music." Based on the constructed tree, the related preferences to classical music are found. It allowed the dissemination and promotion of the new performances and shows in the additional target group and extended the audience. An actual task of developing methods of web profile establishment is used for music preferences analysis. Random forest and KNN are the best predictors for chosen dataset in this study. The feature selection is organized based on random forest. This method allows to fond the most regularly preferences to classical music.