Multi-layer Perceptron Approach for Diabetes Risk Prediction using BRFSS Data

The escalating prevalence of diabetes worldwide, affecting approximately 425 million individuals by 2020, underscores the urgent need for accurate predictive models to address this pressing public health challenge. In this study, we investigate the efficacy of Multi-layer Perceptron (MLP) models with varying units across two hidden layers—specifically, configurations with 5, 10, 15, and 20 neurons per layer. Leveraging the BRFSS dataset for diabetes risk prediction, our meticulous evaluation of these configurations unveils compelling insights. Notably, the MLP model attains its peak accuracy of approximately 84.91% with 5 neurons in Hidden Layer 1 and 10 neurons in Hidden Layer 2. These findings underscore its potential for robust predictive modeling. Furthermore, despite minor fluctuations, our analysis highlights the model’s reliability and effectiveness in achieving high predictive accuracy. This research offers valuable contributions to the field of diabetes risk prediction, providing insights into the optimal configuration of MLP models for accurate and reliable predictions, thus aiding in the development of improved diagnostic tools and preventive strategies to mitigate the global burden of diabetes.