A Crow Particle Swarm Optimization Algorithm with Deep Neural Network (CPSO-DNN) for High Dimensional Data Analysis

Diagnosis of any disease at its early stage correct treatment of the patients is necessary to save productive lives. Many techniques are adopted by different researchers to diagnose the disease at an early stage, but none of them are suitable due to the course of dimension dilemma. To differentiate the gap between logical and biological variation between the samples of the considered datasets, feature selection plays an important role. From different research, it is clear that while considering the local optima, PSO algorithm converges prematurely and decreases the diversity of the population. To avoid the limitations Crow Particle Optimization (CPSO) is implemented to identify the featured genes from a high dimensional dataset. The main concepts behind this CPO are related to the bird crow which hides a large number of foods in different places securely and collects it as per the need. To understand performance of the (CPSO-DNN) we have used three different evolutionary searchings like firefly search, elephant search and squirrel search. Gradient descent based Deep neural network with soft-max activation function has been used to maximize the no of feature genes by reducing the low impact features. Simulation results demonstrate that the (CPSO-DNN) exhibits an outstandingly higher accomplishment in terms of accuracy and can be considered as a better classification model as compared to other algorithms.