A Novel Multi-Stage Feature Selection and Classification Model for Accurate Liver Disease Detection

Liver disease is a pervasive and pressing global health challenge whose conditions vary from viral infections, fatty liver disease, to more severe conditions such as cirrhosis and hepatocellular carcinoma. In order to detect liver disease at earliest stages, there is an urgent need of developing an accurate and efficient diagnostic model. Considering this, a novel and highly accurate NeuroOpt model is proposed in this manuscript for detecting and classifying stages of liver disease in patients. The proposed model utilizes Neural Network architecture for classifying disease whose performance is enhanced by self-tuning process through GWOA2 (Grasshopper and Whale Optimization Algorithm) method. Moreover, an effective three-stage Feature Selection method is also used in the proposed work for meticulously choosing vital attributes from given dataset. The key innovation of this work is the development of a dynamic self-tuned NeuroOpt liver disease detection model along with three-stage Feature Selection that not only predicts the presence and absence of liver disease in humans but also classifies them into multiple stages using two datasets instead of only one. Through the utilization of two datasets, the proposed model became dynamic due to its high accuracy rate for both binary and multi-stage classifications respectively. The efficacy of proposed approach is examined and validated on Indian Liver Patient Dataset (ILPD) and Cirrhosis Prediction Dataset (CPD) using MATLAB software for binary and multi-stage disease classifications respectively. Experimental outcomes determine that proposed model is able to detect disease in patients with an accuracy of 97% (binary classification) on ILPD dataset, whereas, it attained an accuracy of 98.9% for multi-stage disease classification of CPD dataset to prove its effectiveness.