Implementation Of the Decision Tree Algorithm in Predicting Delayed Water Bill Payments in A Housing Residence in Bandung

Abstract

The provision of clean water is a fundamental necessity that depends significantly on consumers' prompt payment of bills for its operational sustainability. The administration of a housing complex in Bandung is experiencing significant difficulties with overdue payments, which are affecting its financial and operational stability. The existing strategy for managing delays is reactive, rendering it ineffective at preventing debt accumulation. This research proposes adopting the Decision Tree algorithm as a predictive method to address this issue. This research employs a quantitative methodology that uses predictive techniques, drawing on previous secondary data on customer water bill payments. The technique utilized is CRISP-DM, encompassing phases of business understanding, data understanding, data preparation, modeling, assessment, and implementation. In the data preparation step, missing values are resolved, feature engineering is performed (calculating average monthly payments), irrelevant columns are eliminated, and categorical variables are encoded. The dataset is subsequently split into 80% for training and 20% for testing. The Decision Tree Classifier model is constructed and trained utilizing the processed training data. Model performance is assessed using Accuracy, Precision, Recall, and F1-Score measures, along with Confusion Matrix analysis and feature significance evaluation. The testing results indicate that the model achieves 78% accuracy in forecasting client payment status (On Time or Late). Nonetheless, the model continues to yield satisfactory outcomes as a preliminary alert mechanism for possible delays. This study demonstrates that machine learning can enhance billing efficiency and customer management within the housing complex in Bandung.