Penerapan Algoritma Naive Bayes untuk Memprediksi Keputusan Berlangganan Deposito Berjangka pada Kampanye Pemasaran Langsung

Abstract

Direct marketing campaigns via telephone calls are a key strategy for banks to offer term deposit products. However, the effectiveness of this strategy is often hindered by the uncertainty of customer responses. This study aims to predict customer decisions in subscribing to term deposits by utilizing data mining techniques. The data used is sourced from the UCI Machine Learning Repository which is multivariate, covering demographic attributes, financial history, and campaign interactions. Through data pre-processing stages to handle missing values and class imbalance, this study applies classification models to map potential customer patterns. Experimental results show that the classification model is able to predict non-subscribing customers very well (92.67% precision), but still faces challenges in detecting subscribing customers (35.00% precision). These findings indicate that while the model can help filter marketing targets, further optimization is needed to address data imbalance to improve prediction accuracy in the minority class.