A Neural Network for Pattern Classification using Fuzzy Set Theory

Abstract

The conventional algorithms related to the Artificial Neural Networks (ANN) have some innate shortcomings, similar to the probability of categorizing in native maximum outcome, which possess reduced speed in the learning procedure, thereby contributing to the failure in seeing a productive cell arrangement. To overcome this lacking factor, this given paper proposes a Neural Network Classifier (NNC) built combining the features of the Beetle Antennae Search (BAS) formula, termed BASNNC and Fuzzy Set Theory, which is a research and analysis proposal that can deal with problems relating to inconclusive, subjective and vague judgments. To enhance the weights of the NNC, the BAS formula is used. BASNNC consists of a three-layer structure- an input layer, a hidden layer and an output layer. The aim is to develop novel neural network that combines the significant features of Neural Network and the Fuzzy Set Theory into a common network. This process aims to get the efficient result while eliminating the errors. The objective will be to develop the system that achieves high accuracy results with the computational complexity, using the pattern classification. A pattern can be viewed physically or mathematically by the application of algorithms. To find results for a given set of observations, pattern classification is used. Quite differing from the normal technique employing the concept of gradient descent, the differences between the hidden and the output layers area are enhanced by the BAS formula, that successfully improves the process and gets the desired results. The domain used in the procedure is Artificial Intelligence (AI).