Swarm optimized fuzzy reasoning model (SOFRM) for diabetes diagnosis

Early diagnosis of diabetes is important as it reduces the chances of related complications to arise. Several clinical factors are taken into account for reaching conclusion regarding presence or absence of the disease in a given case. However, the exact relationship between these factors and the incidence of disease is not known. Moreover, there is no general consensus regarding relative importance of these factors in determining the disease. Classification systems that rely on such factors tend to be computationally complex due to large number of factors and associations. The aim of this paper is to employ succinct yet effective clinical rules for diagnosis of diabetes. The proposed Swarm Optimized Fuzzy Reasoning Model (SOFRM) employs feature selection for selecting the most discriminative features for diagnosis. The selected features are embedded in a Fuzzy Rule Base with the aim of tolerance to imprecisions in feature measurements and individual fluctuations. Further, the fuzzy rule base is optimized using Swarm Intelligence to achieve highest possible accuracy level. Experimental results demonstrate that SOFRM gives comparable or better accuracy in diabetes diagnosis than many other state-of-art machine learning approaches.