Enhancing stacked deep neural network with multi objective optimization using chaotic grasshopper algorithm.
Abstract
Abstract - A novel chaotic Grasshopper optimized Deep Neural Network architecture (CGDNN) is constructed by stacked ensembled methodology with multi-objective optimization for heart disease prediction in diabetes patients in this study. Constructing a DNN architecture by combining multiple Neural Networks using ensembled methods will improve the accuracy of the model. A novel metaheuristic swarm-optimization method known as the Adaptive grasshopper Algorithm which is improved using a chaotic mathematical model is used as the optimizer in this proposed work. The Chaotic Adaptive Grasshopper Optimization mimics the life cycle and food searching pattern of grasshopper insects for solving complex scientific problems. With the rapid advancements in Medical diagnosis, a lot of Neural Network models have been developed to find the right diagnosis. But still, local optima problem is a challenging issue because accuracy is an essential factor in medical diagnosis. In this proposed work, the stacked DNN model is optimized by the CGOA with Multi-objective optimization with MSE, and CC is taken as the objective function. It is found that after optimization the proposed Model shows a significant increase in the performance Precision 0.6, Recall 11.3, Specificity 5.1, F1 Score 6.3, and Accuracy 10.68.