An Approach towards prediction of Diabetes using Modified Fuzzy K Nearest Neighbor

Efficient classification of Diabetes plays a major part in the recognition of the critical condition of a person's health. This will help in providing a quick action of the relevant issues concerning Diabetes. This paper deals with metrics that measure the performance of divergent classification methods like Support Vector Machine (SVM), Naïve Bayes and K Nearest Neighbor (KNN). But the performance and efficiency of these methods are not always satisfactory. To solve the concern of coherent classification for detection of critical conditions for Diabetes, a Modified Fuzzy K Nearest Neighbor classifier has been proposed. The behavior of the Fuzzy KNN classifier depends upon two variables: the occurrence of nearest neighbor classes (K) and the fuzzy coefficient (M). The Gaussian membership function is incorporated here to create a fuzzy set from a crisp set. Different datasets of diabetes are chosen in this work for evaluating the performance of Modified Fuzzy KNN. After analyzing the results, it is observed that the performance of the proposed classification method is better than other classical approaches.