<![CDATA[Accurate population prediction is essential for development planning, particularly in multiethnic countries such as Malaysia, yet conventional statistical approaches generally do not explicitly incorporate ethnic composition. This study aims to develop a population prediction model for Malaysia using a Random Forest Regressor that accounts for demographic factors and the composition of six ethnic groups. Historical data from 16 Malaysian states spanning 1981–2025 (673 state-year observations) were processed through feature engineering, including population lag variables, growth rate, and ethnic proportions, yielding 21 independent variables. The model was trained using a time-aware train-test split (80% training, 20% testing) and validated with 5-fold time series cross-validation. Results on the test set (2020–2025) showed strong performance, with R² = 0.9769, RMSE = 259 thousand people, MAE = 97 thousand people, and sMAPE = 3.17%. Feature importance analysis revealed that the previous year’s population (pop_lag1) and gender composition were the most dominant predictors. Projections indicate Malaysia’s population will reach approximately 38.5 million by 2030. These findings demonstrate that Random Forest with demographic-ethnic features can produce more accurate population predictions, providing a data-driven policy support tool for the Malaysian government.]]>
Copyrights © 2025
