Forecasting rainfall distribution using artificial neural networks for Johor rivers

Abstract

The study is conducted to forecast the rainfall distribution in the areas around Johor, Malaysia. Although there are many other factors, we will be using the rainfall distribution factor only. The forecasting method that is going to be used in this study is the Artificial Neural Networks (ANN) which will be trained using back propagation learning algorithm. To produce the best model, several propagation models will be constructed in the algorithm. The value of learning rate parameter and momentum parameter will also be used and constantly changed based on the number of hidden nodes. The data is prepared and filtered using data pre-processing. Data pre-processing includes data cleaning, normalisation, transformation, feature extraction and selection. The product of data pre-processing is the final training set. At the end of the experiment, the best model was selected and the strength of the relationship of each model based on their activation functions that have been used was compared. The result of the model produces the minimum error value and has a stronger relationship between the actual data value and forecast data value is the best model among the best.