Prediction of Solar Radiation using Hybrid Discriminant-Neural Network

Abstract

A timely and accurate prediction of solar radiation results in proper plant growth, seed germination and stages of flowering and fruiting. Neural Network is becoming popular in designing predictive models. However, issues like importance of variables and long training process has limited its accuracy. The objective of this study is to explore the performance of predictive model by integrating neural network with traditional step-wise discriminant analysis forming a hybrid model. The inclusion of selected features from discriminant analysis to the neural network will improve the accuracy of the designed predicted model. The paper also examines that the hybrid approach outperforms the neural network by selecting different architecture of neural network.