Web Based Software for Back Propagation Neural Network withWeight Decay Algorithm

摘要

Artificial Neural Networks (ANNs) are non-linear structures used for prediction and classification problems. ANNs identify and learn correlated patterns between input data sets and corresponding target values. Trained ANNs are used to predict the outcomes of independent variables. Over fitting and under fitting are two major problems that may arise in ANNs. When two or more predictor variables in a model are highly correlated, called as multi-collinearity, they provide redundant information about the response and leads to overtraining. This problem is handled by using ANN with weight decay algorithm. Many software are available for analyzing the data using ANN but either they are very expensive or difficult to use. This study describes a web based software for back propagation neural networks with weight decay algorithm. This software is useful for statisticians and researchers implementing ANNs for various data mining task and facing non-convergence problem.