AI-Driven Rainfall Prediction: The Role of Data Normalization in ANN Performance

This study explores the application of artificial neural networks (ANN) for predicting average monthly rainfall in Isfahan using data from 1951 to 2014. The focus is on the preprocessing of input data, a crucial step often overlooked, which significantly influences the performance of neural networks. Several data normalization techniques, including logarithmic transformation, outlier removal, and homogenization, were applied to enhance the accuracy of predictions. The study employed a multi-layer perceptron (MLP) architecture to model the data, with inputs derived from the average rainfall of the previous four years. The results demonstrate that normalized and homogenized data significantly improve the network's predictive accuracy, achieving a correlation coefficient (R) of up to 0.97 and a mean square error (MSE) as low as 0.04. These findings underscore the importance of proper data preprocessing in ANN applications for reliable climate prediction.

AI-Driven Rainfall Prediction: The Role of Data Normalization in ANN Performance Keywords:

Artificial Neural Networks (ANN) , Data Normalization , Multi-Layer Perceptron (MLP) , Logarithmic Transformation