KOMPARASI TIGA MACHINE LEARNING KLASIFIKASI UNTUK PREDIKSI CHURN

AKMAL, MUHAMMAD HIBATUR (2024) KOMPARASI TIGA MACHINE LEARNING KLASIFIKASI UNTUK PREDIKSI CHURN. S1 thesis, Universitas Mercu Buana Jakarta.

	Text (HAL COVER) 01 Cover.pdf Download (458kB)
	Text (ABSTRAK) 02 Abstrak.pdf Download (34kB)
	Text (BAB I) 03 Bab 1.pdf Restricted to Registered users only Download (132kB)
	Text (BAB II) 04 Bab 2.pdf Restricted to Registered users only Download (325kB)
	Text (BAB III) 05 Bab 3.pdf Restricted to Registered users only Download (159kB)
	Text (BAB IV) 06 Bab 4.pdf Restricted to Registered users only Download (777kB)
	Text (BAB V) 07 Bab 5.pdf Restricted to Registered users only Download (58kB)
	Text (DAFTAR PUSTAKA) 08 Daftar Pustaka.pdf Restricted to Registered users only Download (102kB)
	Text (LAMPIRAN) 09 Lampiran.pdf Restricted to Registered users only Download (237kB)

Abstract

High bank customer churn or the movement of customers from one bank to another can be detrimental to the bank. Customers as valuable assets owned by the company must be maintained well. Predicting customer churn can be a solution to this problem, because by knowing the characteristics of customers who will churn, banks can take preventive action so that customers do not churn and at the same time reduce the churn rate. The technology and collection of customer data currently available in banks can be used as tools and materials to predict churn using data analysis or data mining. Classification machine learning algorithms such as KNN, Decision Tree, and XGBoost are reliable enough to compare model results in terms of accuracy, precision, recall, and AUC scores. The CRISP-DM method which is commonly used in data mining plays an important role as a research method. A dataset that has been well prepared, then processed using log transformation, standardization, and class imbalance handling techniques is able to improve each model metric value as suggested in previous research. The results of this research show that XGBoost without feature selection achieved the highest accuracy and precision values with values of 86% and 75% compared to KNN and Decision Tree. Further research can increase the low recall and F1 scores in each model in this study. Keywords: Churn, Decision Tree, K-Nearest Neighbor, Machine Learning, XGBoost. Tingginya bank Customer churn atau pindahnya nasabah dari satu bank ke bank lain dapat merugikan bank. Nasabah sebagai aset berharga yang dimiliki oleh bang harus bisa dipertahankan dengan baik. Prediksi customer churn dapat menjadi solusi permasalahan ini, karena dengan mengetahui ciri nasabah yang akan churn, bank dapat melakukan tindakan preventif agar nasabah tidak churn sekaligus mengurangi tingkat churn. Teknologi dan kumpulan data customer yang ada di bank saat ini dapat digunakan menjadi alat dan bahan dalam memprediksi churn menggunakan analisis data atau data mining. Algoritma machine learning klasifikasi seperti KNN, Decision Tree, dan XGBoost cukup handal untuk melakukan perbandingan hasil model dilihat dari nilai akurasi, precision, recall, dan auc score. Metode CRISP-DM yang biasa digunakan dalam data mining berperan penting sebagai metode penelitian. Dataset yang telah dipersiapkan dengan baik, kemudian diproses menggunakan teknik log transformation, standardization, dan class imbalance handling mampu memperbaiki setiap nilai metric model seperti yang sudah disarankan pada penelitian sebelumnya. Hasil dari penelitian ini menunjukkan bahwa XGBoost tanpa feature selection meraih nilai akurasi dan precision terbesar dengan nilai 86% dan 75% dibandingkan KNN dan Decision Tree. Penelitian selanjutnya dapat meningkatkan nilai recall dan F1 score yang rendah pada tiap model di penelitian ini. Kata kunci: Churn, Decision Tree, K-Nearest Neighbor, Machine Learning, XGBoost

Actions (login required)

View Item