Rainfall Prediction using Artificial Neural Network with Forward Selection Method

The weather has become an important part of people's daily activities; therefore, many people need faster, more complete, and more accurate information about its condition. Accurate weather predictions can be used to solve problems arising from weather effects. Compared to other methods, the Artificial Neural Network (ANN) method is deemed more efficient in fast computing and is able to handle unstable data in terms of weather forecast data. However, ANN has limitations in studying classification patterns if the dataset has large data and high dimensions. To manage this limitation, a feature selection method is needed to enable the ANN to produce accurate predictions. Several experiments were carried out to obtain the optimal architecture and produce accurate predictions. The proposed method only reduces the accuracy value to less than 1% and the loss value to less than 0.01 in both tested datasets. With these results, it can be said that the proposed method is feasible to be used as an improved method for the ANN algorithm.