Predicting Consumer Payment Preferences Using Naïve Bayes: A Supervised Learning Approach for Behavioral Insight

Abstract

Abstract

In the age of digital transformation, understanding consumer payment preferences is critical for designing effective financial systems and service strategies. This study proposes a supervised learning approach using the Naïve Bayes Classifier to predict consumer payment methods—cash or digital—based on key demographic and behavioral attributes, including gender, age, and transaction history. A synthetic dataset comprising 500 records was developed to reflect real-world consumer profiles. The methodology involved data preprocessing, model development, and performance evaluation using standard classification metrics: confusion matrix, precision, recall, F1-score, and AUC-ROC. The resulting model achieved 93.60% accuracy, with a precision of 93.25% for digital payment prediction and a recall of 93.92% for cash payment classification. The AUC-ROC score of 0.986 indicates excellent discriminative performance. These findings demonstrate the practical utility and efficiency of the Naïve Bayes algorithm in capturing behavioral patterns from limited input features. The approach is relevant for developing intelligent recommendation systems, supporting risk assessment, and advancing digital financial literacy—particularly in emerging sectors such as education, including Islamic boarding schools (pesantren), which are increasingly integrating digital financial tools.