Web-Based Boarding House Management Information System with Dashboard Analytics and Multiple Linear Regression for Rental Income Prediction

Small boarding-house businesses commonly rely on receipts and annual notebooks to record tenants, rental periods, payments, unit availability, and income. Such practices can support daily operations, but they make historical retrieval, operational monitoring, and decision support increasingly difficult as the number of records grows. This study develops a web-based boarding house management information system that integrates operational data management, dashboard analytics, and rental-income prediction for Surapati Boarding House in Bandung, Indonesia. The application was developed with Laravel 12, PHP, and MySQL, while the predictive model was trained offline in Python using scikit-learn. Because the original tenant and payment records contain private information, a simulated historical dataset was constructed from the property’s operational characteristics. The dataset contains 54 monthly observations from January 2022 to June 2026. The independent variables were occupied Type A units, occupied Type B units, and payment arrears, while monthly rental income was the dependent variable. After cleaning currency-formatted values and removing invalid rows, the data were divided into 43 training observations and 11 testing observations. The resulting multiple linear regression model was integrated into the Laravel analytics page through its estimated intercept and coefficients. The system provides authentication, unit and tenant master data, rental transactions, payment records, operational summaries, due-date reminders, and an analytics interface that compares actual and predicted income. Model evaluation produced an R-squared value of 95.85%, a mean absolute error of IDR 250,715.83, a root mean squared error of IDR 303,605.01, and a mean absolute percentage error of 1.39%. These results indicate close agreement between predicted and simulated actual income. However, the findings are limited to the simulated case-study dataset and require validation using anonymized real operational data before broader deployment.