A Conceptual Machine Learning Approach to Rainfall Pattern Prediction in Umuahia Metropolis Using Multilayer Perceptron

The erratic nature of rainfall in Umuahia metropolis of Abia State, Nigeria due to the constant variations in atmospheric conditions, results in extreme weather conditions such as drought and flooding which pose dire consequences to human beings and the environment resulting in loss of lives, damage to agricultural produce and vital infrastructure. This study seeks to present a conceptual approach using a machine learning algorithm to support rainfall pattern prediction. A comprehensive reviews of related works was carried out on Artificial Neural Network (ANN), Support Vector Machine (SVM), Multilinear Regression (MLR) and Extreme Gradient Boosting (XGBoost) applications in prediction. The min-max normalization technique was deployed to render the dataset in a common normalized data range. A 4-10-1 architecture of a Multilayer Perceptron (MLP) was designed with four nodes at the input layer, ten nodes at the hidden processing layer, and one node at the output layer for rainfall pattern prediction. Implementation of this study with real data and its comparison with other machine learning algorithms are highly recommended for further study in this domain.