To diagnose diabetes disease at an early stage is quite a challenging task due to complex inter dependence on various factors. There is a critical need to develop medical diagnostic decision support systems which can aid medical practitioners in the diagnostic process. For detection of diabetes at an early stage, data mining association algorithm is being used. Detecting the cause of diabetes is very important in order to stop diabetes. The data set is taken from Nirmay Diabetes Super Speciality Center repository containing total instances 900 and approximately 30 attributes of type 2 diabetes mellitus. Proposed method explores step-by-step approach to help the health doctors to explore their data and to understand the discovered rules better. The discovery of knowledge from medical databases is important in order to make effective medical diagnosis. Most of the rules focus on the improving mining efficiency but, for the medical research, mining efficiency is not the most important factor, and there is a need to find more useful association rule and abandon more useless association rule. We proposed a modified equal width binning interval approach to discretizing continuous valued attributes. The approximate width of desired intervals is chosen based on the opinion of medical expert and is provided as input parameter to the system. First we have converted numeric attributes into categorical form based on above techniques. FP-growth association mining is used to generate rules which are useful to identify general association in the data.