Optimized design of hybrid genetic algorithm with multilayer perceptron to predict patients with diabetes

Diabetes is a set of long-term metabolic issues characterized by high blood glucose levels over a drawn-out time. It is a challenging and threatening disease because nearly half of patients who have diabetes do not know that they have it. Data mining techniques have been heavily utilized to predict diabetes in its earliest stage, and therefore can potentially reduce medication costs. Most of the existing research used metaheuristics to optimize neural network parameters or used other data mining algorithms to classify diabetes. This research integrates design of experiments (DOE), genetic algorithm (GA), and multilayer perceptron (MLP) to classify diabetic patients. Two sets of factorial designs were used to determine an optimal set of parameters for the GA. GA is used to optimize the parameters in the MLP, where the MLP acts as a simulator within the GA to provide the fitness evaluation of the solutions. The best resulting solution from DOE-GA-MLP represents the optimized MLP that achieves the highest classification accuracy in diabetes patients. The proposed approach outperforms eight different classification algorithms in terms of accuracy, sensitivity, and F1-score. The developed DOE-GA-MLP model presents a robust predictive tool that can help healthcare professionals identify patients with diabetes in its early stage. Intervention and/or preventive strategies can be implemented in those identified cases to improve patient care and possibly reduce diabetes rates and potential expenditures.

Automated summary

Diabetes is a challenging and threatening disease, and data mining techniques have been applied to predict and classify diabetic patients. This research integrates design of experiments (DOE), genetic algorithm (GA), and multilayer perceptron (MLP) to classify diabetic patients. The proposed approach outperforms eight different classification algorithms and presents a robust predictive tool for early detection of diabetes.