The number of patients that were infected by Diabetes Mellitus (DM) has reached 415 million patients in 2015 and by 2040 this number is expected to increase to approximately 642 million patients. Large amount of medical data of DM patients is available and it provides significant advantage for researchers to fight against DM. The main objective of this research is to leverage F-Score Feature Selection and Fuzzy Support Vector Machine in classifying and detecting DM. Feature selection is used to identify the valuable features in dataset. SVM is then used to train the dataset to generate the fuzzy rules and Fuzzy inference process is finally used to classify the output. The aforementioned methodology is applied to the Pima Indian Diabetes (PID) dataset. The results show a promising accuracy of 89.02% in predicting patients with DM. Additionally, the approach taken provides an optimized count of Fuzzy rules while still maintaining sufficient accuracy.