Air Temperature Forecasting Based on Stacking Machine Learning Model with Multi-Step Time Series

Forecasting air temperature is very influential in various areas of fields. Temperature is an fundamental factor that influences weather. Changes in the temperature in that environment can mark changes in environmental weather. Forecasting the inside temperature and long waiting time determines the planning environment and planning of electrical energy resources. The purpose of this study is to forecast air temperature from the data recorded by the temperature sensor. Therefore, in this work, we propose that temperature forecasting is based on an ensemble of numerous machine learning models to obtain the best model accuracy. We have used the stacking ensemble model to forecast air temperature. A combination of models to find the best model is needed to build a predictive model. Using the base layer model and the number of base classifier models significantly affects the model prediction error. The best ensemble model accomplished is the combination model (Gradient Boosting Regression-Support Vector Regression) - Gradient Boosting Regression meta-model with a MAPE value of 0.0096.