ANALISIS PENGARUH FITUR EMOSI MUSIK TERHADAP POPULARITAS LAGU SPOTIFY MENGGUNAKAN RANDOM FOREST DAN SHAP

Abstract

The rapidly growing digital music industry, enabled by the Spotify platform, enables in-depth analysis of the emotional factors that influence a song's popularity. This study aims to identify significant emotional features, analyze their influence on the number of streams, and develop a data-driven model for predicting song popularity. The research method is an experimental artificial intelligence (machine learning) approach that applies the Random Forest Regression algorithm combined with the Shapley Additive Explanations (SHAP) interpretability method. The research stages include collecting datasets from Kaggle, data preprocessing, model training, and evaluation using the Coefficient of Determination (R²), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The results show that valence, danceability, acousticness, and energy have the greatest influence on the number of streams, while instrumentalness has the least. Based on the Shapley Additive Explanations analysis, high emotional values, valence, and danceability contribute positively to song popularity. This study concludes that combining the Random Forest algorithm with the Shapley Additive Explanations method is effective for predicting and understanding the emotional factors that determine song success on Spotify.

Industri musik digital yang berkembang pesat melalui platform Spotify memungkinkan analisis mendalam mengenai faktor emosional yang memengaruhi popularitas sebuah lagu. Penelitian ini bertujuan untuk mengidentifikasi fitur emosional yang signifikan, menganalisis pengaruhnya terhadap jumlah stream, serta membangun model prediksi popularitas lagu menggunakan pendekatan berbasis data. Metode penelitian yang digunakan adalah pendekatan eksperimental kecerdasan buatan (machine learning) dengan menerapkan algoritma Random Forest Regression yang dipadukan dengan metode interpretabilitas SHapley Additive Explanations (SHAP). Tahapan penelitian meliputi pengumpulan kumpulan data (dataset) dari Kaggle, pra-pemrosesan data, pelatihan model, serta evaluasi menggunakan metrik Coefficient of Determination (R²), Mean Absolute Error (MAE), dan Root Mean Square Error (RMSE). Hasil penelitian menunjukkan bahwa fitur valence, danceability, acousticness, dan energy memiliki pengaruh paling signifikan terhadap jumlah stream, sementara fitur instrumentalness tercatat memiliki pengaruh paling kecil. Berdasarkan analisis SHapley Additive Explanations, nilai emosional yang tinggi pada fitur valence dan danceability berkontribusi positif terhadap popularitas lagu. Penelitian ini menyimpulkan bahwa kombinasi algoritma Random Forest dan metode SHapley Additive Explanations efektif dalam memprediksi serta memahami faktor-faktor emosional yang menentukan kesuksesan lagu di platform Spotify