In recent years, the number of diabetic patients has been increasing rapidly all over the world. Diabetes cannot be completely cured once it develops, so it is important to prevent diabetes before it develops. To prevent the onset of diabetes, it is necessary to avoid prolonged hyperglycemia after meals. In this paper, we propose a self-management system to help users prevent diabetes. The system monitors blood glucose levels in real time, calculates foods to be reduced taking into account the user's preferences, and presents them to the user as soon as the system predicts that the planned diet will cause high blood glucose. We designed and conducted two experiments to show the effectiveness of the proposed system. Experiment 1 was to construct and evaluate a model for predicting blood glucose levels two hours later. Experiment 2 was to evaluate the degree of satisfaction with the food and recommendations, and the acceptability of the recommendations by participants who actually used the proposed system. The results of Experiment 1 showed that the constructed model was able to predict blood glucose levels with an RMSE of 7.66 and MAE of 4.66. As a result of Experiment 2, we found the recommended intake was more acceptable if it reflected the user's preferences.