Disease Prediction Using a Modified Multi-Layer Perceptron Algorithm in Diabetes

Abstract

This paper presents an adaptation of the Multi-Layer Perceptron (MLP) algorithm for use in predicting diabetes risk. The aim is to enhance the accuracy and generalizability of the model by incorporating preprocessing techniques, dimensionality reduction using Principal Component Analysis (PCA), and improvements in optimization and regularization. Several factors, including glucose level, pregnancy, blood pressure, and body mass index, are taken into account when analyzing the PIMA Indian Diabetes dataset. Modern optimization methods, dropout regularization, and an adaptive learning rate are incorporated into the modified MLP model to fine-tune the model's weights and boost its predictive abilities. The effectiveness of the modified MLP algorithm is evaluated by comparing its performance with baseline machine learning methods and the original MLP algorithm in terms of accuracy, sensitivity, and specificity. The results of this study can improve the quality of healthcare provided to people at risk for developing diabetes and thus contribute to the development of better prediction models for the disease.