Investigating the Relationship Between Music and Human Psychology Through Big Data Analytics

Human emotional responses to music are commonly described using the valence–arousal model, which represents emotions according to their positivity (valence) and intensity (arousal). Predicting these dimensions from musical features enables researchers to better understand the relationship between music and human psychology. This study proposes a scalable, data-driven framework that integrates acoustic feature extraction, statistical feature engineering, exploratory data analysis (EDA), and supervised machine learning to predict valence and arousal from music. The Database for Emotional Analysis in Music (DEAM) dataset is used because it contains listener-annotated valence and arousal ratings that reflect perceived emotional responses to musical stimuli. A comprehensive feature set is extracted, including Mel-Frequency Cepstral Coefficients (MFCCs), spectral features, and energy-based features. A LightGBM regression model is employed to capture nonlinear relationships between audio characteristics and emotional dimensions. Experimental results demonstrate strong predictive performance, achieving Root Mean Squared Error (RMSE) values of 0.1492 for valence and 0.1416 for arousal, with Mean Absolute Error (MAE) values of 0.1108 and 0.1085, respectively. The findings demonstrate the effectiveness of the proposed framework for modeling music-induced emotional responses and highlight its potential applications in personalized music recommendation, emotion-aware multimedia systems, and music therapy.