Predictive Analysis of Obesity Cases Using Linear Regression and Weighted Moving Average

Abstract

Obesity remains a persistent public health problem in Indonesia, including in Bandung City, where the local Health Office has periodically collected obesity screening data but has mainly used it for descriptive reporting rather than predictive analysis. This study analyzes and forecasts the monthly number of obesity cases in Bandung City using Linear Regression and Weighted Moving Average (WMA), and compares their accuracy. The WMA weighting scheme for a three-period window was determined through experimentation, and WMA was selected as the comparison method after being benchmarked against Simple Moving Average and Exponential Smoothing on testing data. Twenty monthly aggregate observations from October 2024 to May 2026 were split into 80% training and 20% testing data. Model performance was evaluated using Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Root Mean Squared Error (RMSE). Results show that Linear Regression outperformed WMA on all three metrics, with MAE of 1,419.08 versus 1,448.56, MAPE of 17.00% versus 19.64%, and RMSE of 1,525.29 versus 1,788.29. Forecasts for the following two months produced higher estimates from Linear Regression than from WMA. These findings provide an initial step toward the predictive use of local health surveillance data to support evidence-based obesity prevention planning.