Comparing Gradient Descent and Genetic Algorithm for Optimization In Regression Analysis For Handling Nonlinearity In Liver Cirrhosis: A Survey

The development of a noninvasive diagnostic paradigm for digestive illnesses is the primary focus of the work being done in clinical research at the moment. We want to demonstrate the efficacy of both the BP-ANN algorithm and linear regression in the diagnosis of GI illnesses by contrasting their activation functions and data structures. PRISMA criteria were followed in the reporting of the systematic review that we conducted. We examined comparable papers in seven different online academic sources for the purpose of comparing the accuracy of diagnosis utilizing Our principal interests in BPANN and linear regression. Regardless of whether or not a given study met the inclusion requirements, we were able to acquire features, patient count, input/output marker, diagnostic accuracy, and results/conclusions related to comparison. Since just nine articles met our criteria, we chose to include them in our analysis. Our study is one of eight that found the BP-ANN model to be more effective than linear regression in proposing the course of disease using AUROC scores. One study found that linear regression was more effective than BPANN in spotting early signs of colorectal cancer. While both the BPANN method and the linear regression technique showed promise for fitting the diagnostic model, the former provided more convincing evidence of superior prediction accuracy when applied to a model of noninvasive diagnostics for gastrointestinal disorders. We compared the BP-ANN and linear regression activation functions and data formats to see which would work best for fitting the diagnostic model. The results revealed that BPANN was a complete recommendation method.