A radii-based evolutionary algorithm is applied in solving a difficult classification problem concerning diabetes diagnosis. The algorithm was designed to treat multimodality and has recently successfully been applied in the optimization of several multimodal functions. The medical problem to be solved is to predict the diagnosis – either diabetic or not – for a set of patients, given some personal and medical conditions. Proposed algorithm gives a high accuracy of prediction and thus provides a good means of understanding the factors that doctors consider when diagnosing diabetes and a way of checking the consistency of decision making. A clear advantage over other computational techniques is that, besides the outcome for patients in the test set, the algorithm also provides simple rules that led to that decision.