Diabetes has been recorded already affects 424 million people and it was predicted to increase to more than 600 million by 2045. This encourages many researchers to develop a prediction model that can help minimize this phenomenon by identifying the factors that can improve the probability of having diabetes. However, most researchers only focus on increasing accuracy values because we cannot use the same evaluation metrics for all case studies. This research aims to build a model that has high sensitivity and is easy to interpret through explainable machine learning. The algorithms used are Decision Tree, Logistic Regression, Random Forest, and XGBoost. This research also experiments with resampling data to minimize the risk of bias caused by imbalanced class proportions. The research results show that XGBoost can produce a sensitivity value of 73%. The application of explainable machine learning also shows that the factors that increase diabetes are glucose, Body Mass Index (BMI), and age. The model also indicates that high levels of glucose in the body can improve the probability of having diabetes.