Rainfall Prediction using Neural Network Model with Autocorrelation Function and Correlation Analysis

Weather conditions are one of the vital factors that influence aircraft takeoff activities at airports. This is because weather conditions can affect the lift force of an aircraft. The weather conditions in Indonesia, as a country located in a tropical region with a vast ocean area and surrounded by two continents, make it difficult to accurately predict the weather due to its highly dynamic nature. This research proposes the utilization of pre-processing methods, specifically ACF and correlation analysis, on past weather observation data as input to predict rainfall for the next day in the vicinity of Ahmad Yani Semarang Airport. This method is crucial for analyzing time-series data. ANN with backpropagation based model is used in this study to predict rainfall. Using the same model configuration, we compared the output result from the dataset before applying ACF and correlation analysis and after applying it. It was found that by applying ACF and correlation analysis to our dataset, we managed to achieve a better prediction of rainfall by reducing the RMSE and MAE values. We had initial RMSE and MAE values of 14.8 and 12.6, respectively, before employing ACF and correlation analysis on our dataset. However, by implementing these techniques, we were able to decrease the RMSE to 7.1 and the MAE to 6.5.