Improving Temperature Prediction Accuracy Using Stacking Ensemble Methods in Meteorological Data

Abstract

Accurate temperature prediction is crucial for weather forecasting, influencing sectors such as agriculture, energy, and disaster management. This study aimed to improve daily temperature prediction using stacking ensemble methods on weather data from Pontianak (2021-2024). The research compared the performance of individual models, Linear Regression and Random Forest, and demonstrated how stackingwhich combines multiple modelscould enhance prediction accuracy. Stacking was implemented with Linear Regression and Random Forest as base learners, while a Linear Regression model served as the meta-learner to integrate the predictions from the base models. The results showed that the stacking model outperformed both individual models in terms of Mean Squared Error (MSE) and Mean Absolute Error (MAE), leading to more accurate predictions. While stacking improved accuracy, it introduced greater computational overhead compared to single-model approaches. This trade-off between accuracy and computational efficiency must be considered for real-world applications. The study demonstrated the effectiveness of stacking in enhancing temperature prediction accuracy, offering insights into how ensemble methods can improve weather forecasting.