Application of Multiple Linear Regression Models for prediction of rice production yields in Central Lampung

Abstract

Rice production is a crucial component of agricultural sustainability and food security in Indonesia, particularly in Central Lampung. This study aims to analyze the influence of planting area and harvested area on rice production using a multiple linear regression approach. The analysis employs secondary time-series data and applies an ordinary least squares (OLS) method with a logarithmic transformation of the dependent variable to address heteroskedasticity issues. Descriptive statistics and classical assumption tests, including normality, multicollinearity, heteroskedasticity, and autocorrelation tests, were conducted to ensure model validity. The results indicate that harvested area has a statistically significant positive effect on rice production, while planting areas shows a negative but statistically insignificant effect. The regression model demonstrates strong explanatory capability with an R-squared value of 81.27% and is statistically significant based on the F-test. Model evaluation using in-sample error metrics yields a Mean Absolute Error (MAE) of 19,344.89, a Root Mean Squared Error (RMSE) of 46,738.41, and a Mean Absolute Percentage Error (MAPE) of 48.20%, indicating that the model effectively captures general production trends but has limited accuracy for precise quantitative forecasting. These findings suggest that harvested area plays a dominant role in determining rice output, while further improvements in predictive performance may be achieved by incorporating additional explanatory variables and exploring alternative modeling techniques.