A Novel GA-NN Approach for Feature Optimization and Classification

Abstract

With the emergence of artificial intelligence in the field of computer science the utilization of electronic machines are displaced from calculation to executing intellect. The advancement in technology triggered the experts to adhere the psychology and biology with electronics. With the practice of artificial intelligence the machine attempt to classify data into already categorized classes which are analyzed according to the experience acquisitioned in the training session.In this paper a hybrid model i.e. `GA-LDA-NPR' is used for optimization and classification of datasets. The datasets used for experiment are cancer, iris and wine. The optimization take place using GA-LDA and the optimized data set proceed to NPR tool for further training and classification.After observing the classification results a final conclusion is made that during optimization as the number of selected features increases the classification accuracy will increase up to a limit and if we continuously increases the number of selected features the classification rate starts to degrade. The outcome of this analysis is that if the number of parameters increases continuously the problem of over fitting arises which reduces the performance of classifier.