Analysis of Submitting Payments at The End of The Year in the PDAM Budget Management System Using Random Forest Classifier

Abstract

Submission of payments to providers as part of budget absorption in the budget management system at PDAM Surabaya is an important stage in carrying out the company's activity program. The classic problem is the tendency for payment applications to pile up at the end of the year or in the fourth quarter, resulting in several findings that have not been completed or canceled. Automatically there is also the risk of an increase in cash flow, an increase in the workload in the finance department and a lack of quality spending or company investment as well as vendor difficulties. This study aims to model the random forest classifier and analyze the tendency of accumulation of payment requests in the fourth quarter. The research process starts with data collection and processing using the Machine Learning method, in this case the development of the Random Forest Classifier model, data testing, and analysis of test results. The performance evaluation of the model shows very good results with near-perfect scores (1.0) for all evaluation metrics used: accuracy, precision, recall, and F1- scores. The AUC (Area Under the Curve) value on the testing data is also very high, although slightly lower than that of the training data. Overall, the evaluation results show that a model that has been trained using training data can very well generalize and predict new data in data testing. This indicates that the model has good performance and can carry out classifications with high accuracy