<![CDATA[Multivariate time series forecasting is essential for understanding the interrelationships among weather parameters. This study aims to develop a multivariate forecasting model using a hybrid Vector Autoregressive Integrated Moving Average (VARIMA)–Artificial Neural Network (ANN) approach with the backpropagation algorithm, applied to weather data from Gorontalo Province over the 2015–2023 period, including air temperature, humidity, and wind speed. The data were divided into training data (2015–2021) and testing data (2022–2023). The VARIMA model was employed to capture the linear component, while the residuals from the VARIMA model were subsequently modeled using ANN to capture nonlinear patterns. The order of the VARIMA model was determined based on the smallest Akaike Information Criterion (AIC) value, while model performance was evaluated using Mean Absolute Percentage Error (MAPE). The results indicate that the best-performing model is VARIMA(5,1,1)–ANN(18,9,3), with MAPE values of 1.32% for air temperature, 20.54% for humidity, and 21.96% for wind speed. These findings suggest that the hybrid VARIMA–ANN approach provides good forecasting performance and has the potential to serve as an alternative method for multivariate weather forecasting. ]]>
Copyrights © 2026
