SALM-NARX: Self Adaptive LM- based NARX model for the prediction of rainfall

Prediction of weather has become important as they protect the societal infrastructures and is favorable for farmers. Several techniques have been developed till now to predict the extremity of rainfall. The existing techniques failed to predict the rainfall accurately due to the changing weathers. This paper proposes a prediction model using Nonlinear Auto-Regressive with external input (NARX), which is trained by the proposed Self Adaptive Levenberg-Marquardt (Self Adaptive LM) algorithm. This algorithm is used for selecting the weights optimally. The LM algorithm is enhanced with the learning rate to make it adaptive to yield accurate prediction. The performance of the proposed SALM-NARX is evaluated using two metrics, namely Percentage Root Mean Square Difference (PRD), and Mean Square Error (MSE) and computed as least MSE value with 0.008 and PRD value as 1.721%.