Fuzzy tsukamoto membership function optimization using PSO to predict diabetes mellitus risk level

Abstract

Diabetes Mellitus (DM) is known as the silent killer because the sufferer often goes unnoticed and when it is known, complications usually already occur. The number of people with DM in Indonesia is a lot, which may increase the health costs and may cause trouble for health workers. Therefore, researchers conducted research to make predictions of the risk level of DM. By creating a prediction model the risk of DM can be identified in the early stage. The prediction of DM risk is based on the input variable consisting of age, body mass index and blood pressure (systolic) which are proven to be the basis for determining the risk ratio of DM and the output variable is the level of diabetes risk (low, high). This study uses a combination of Fuzzy Tsukamoto and PSO to predict the risk level of DM. The membership function for Fuzzy Tsukamoto will be optimized with PSO. Membership optimization is expected to increase the prediction results to be more accurate. Based on user data that has been processed using the proposed method, the prediction results are more accurate than the data processed using only Fuzzy Tsukamoto. The MSE value generated between the actual data and the proposed method is 0.012. The resulting MSE value is very small, so this proves the high level of accuracy.