Prediksi Customer Churn Menggunakan Metode CRISP-DM Pada Industri Telekomunikasi Sebagai Implementasi Mempertahankan Pelanggan

Abstract

Fulfilling the needs and desires of customers is a strategy in retaining customers. Among these strategies is a quality product and quality service will also fulfill customer expectations. The purpose of this research is to predict whether customers will become loyal customer or leave the company's services. The research method uses the CRISP-DM (Cross Industry Standard Process for Data Mining) technique and the hypermeter tuning method with the ridge classifier algorithm and the confusion matrix. CRISP-DM is a process model that serves as the base for a data science process with six sequential phases. The data used is 7,043 records. divided into 70:30 data train and data test. From the selection of features, most unsubscribed customers are customers who do not use several services such as VPN, Data Backup, Device Protection, Technician Assistance, TV Streaming, Movie Streaming monthly contracts and use the E-Wallets payment method. Then the results of the accuracy research using the confusion matrix show quite good results with an accuracy of 80.5%, Precision of 85.7% and Recall of 89.8%. As many as 73% or 5,174 continue to use the service and 1,869 or around 27% of customers stop using Telkom Indonesia services.