Studi Komparatif Kinerja LightGBM Dan CatBoost Pada Data Churn Pelanggan Streaming Musik

Royana, Akge Ninov (2025) Studi Komparatif Kinerja LightGBM Dan CatBoost Pada Data Churn Pelanggan Streaming Musik. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21081010097_Cover.pdf Download (1MB)
	Text (BAB I) 21081010097_BAB 1.pdf Download (65kB)
	Text (BAB II) 21081010097_BAB 2.pdf Restricted to Repository staff only until 18 June 2028. Download (909kB)
	Text (BAB III) 21081010097_BAB 3.pdf Restricted to Repository staff only until 18 June 2028. Download (1MB)
	Text (BAB IV) 21081010097_BAB 4.pdf Restricted to Repository staff only until 18 June 2028. Download (1MB)
	Text (BAB V) 21081010097_BAB 5.pdf Download (89kB)
	Text (Daftar Pustaka) 21081010097_DAFTAR PUSTAKA.pdf Download (149kB)

Abstract

The music streaming service industry has experienced rapid growth in recent years. However, intense competition has led to a high rate of customer churn, which directly impacts company revenue. Therefore, the ability to predict customers who are likely to churn is crucial for effective retention strategies. This study compares the performance of two machine learning models, LightGBM and CatBoost, in predicting customer churn in music streaming services. The methodology follows the CRISP-DM framework, using the Streaming Subscription Churn Model dataset from Kaggle. The process includes preprocessing, normalization, and feature engineering, with data splits of 60:40, 70:30, and 80:20 for training and testing. The results show that the CatBoost model with a 70:30 split achieves the best performance, with an accuracy of 84.98%, AUC of 0.9411, recall of 0.85, precision of 0.8559, and F1-score of 0.8533. Based on these results, the CatBoost model proves to be more effective than LightGBM in predicting customers who are likely to churn.

Item Type:

Thesis (Undergraduate)

Contributors: