Comparison of ARIMA boost, Prophet boost, and TSLM models in forecasting Davao City weather data

Abstract

The geography of the Philippines experiences climate variability thus, providing accurate and timely weather forecasts to the population is crucial. Climate forecasts, which are issued and disseminated by government agencies, serve as essential risk management tools. However, the country faces challenges in forecasting, further exacerbated by climate change. Thus, exploring the use of artificial intelligence has emerged as a strategy to enhance weather prediction accuracy. This research focuses on time series forecasting of rainfall, mean temperature, relative humidity, and wind speed weather data using a machine learning approach. Specifically, it aims to compare and identify the most beneficial forecasting models among autoregressive integrated moving average (ARIMA) boost, Prophet boost, and time series linear model (TSLM). It also seeks to evaluate the performance of these models using mean absolute error (MAE), mean absolute percentage error (MAPE), mean absolute scaled error (MASE), symmetric mean absolute percentage error (SMAPE), root mean squared error (RMSE), and R squared (RSQ) metrics. Results showed that the selection of the forecasting model varies based on the specific parameter under consideration, with no hyperparameter tuning in the analysis. For wind speed, ARIMA boost proves to be a favorable choice. At the same time, TSLM demonstrates effectiveness for relative humidity and mean temperature. Both ARIMA boost and TSLM exhibit strong performance for rainfall. Prophet boost consistently ranks as the least-performing model.