Modified Genetic Algorithm and Polynomial Distribution Based Convolutional Neural Network for Asthma Disease Diagnosis

Abstract

Prevention is considered to be the key solution for asthma rather than treatment for the same. Generally, asthma starts initially at early life, hence primary asthma determination in young children is regarded as significant life saving activity. In prevailing scheme, classification necessitates extra time for task completion as well disease prediction cannot be accomplished precisely using enormous samples for evaluation. Modified Genetic Algorithm and Polynomial distribution based Convolutional Neural Network (MGA+PCNN) algorithm is greatly utilized for mitigating these issues and this methodology encompasses key steps such as normalization, pre-processing, feature selection and classification process. Improved Variance Stabilizing Normalization (IVSN) is exploited for dataset accuracy improvement. Pre-processing is achieved using logistic regression clustering algorithm which helps in increasing asthma patient’s attributes significance and these pre-processed features are considered for feature selection process. MGA algorithm is suggested for highest f-measure features selection from specified dataset via best global and local fitness values. Subsequently, PCNN algorithm is deployed for accurate outcome classification for asthma disease identification. It is thereby validated that suggested MGA+PCNN algorithm offers improved True Positive Rate (TPR), False Positive Rate (FPR), F-measure and accuracy than prevailing approaches.