Predicting hepatitis presents a significant challenge in the field of medical data analysis. Healthcare professionals, including doctors and hepatologists, require automated tools to assist in decision-making and to differentiate between healthy and infected liver conditions. The Mixed Radial Basis Function Neural Network (MRBFNN) is a feedforward artificial neural network that uses radial basis functions as activation functions in the hidden layer. In this paper, we propose a model designed to optimize the selection of radial basis functions, centers, variances, output weights, and architecture for accurate hepatitis prediction. The optimization problem is framed as a mixed-variable optimization task with linear constraints. To address this challenge, the authors propose employing a Differential Evolution (DE)-based approach. This methodology is applied to hepatitis prediction, utilizing the optimized parameters to significantly improve the model’s overall accuracy and predictive performance.