Adaptive Divergence Weight Firefly Algorithm (ADWFA) with Improved K-Means Algorithm and Adaptive Neuro Fuzzy Inference System (ANFIS) for Type 2 Diabetes Mellitus Prediction

M. Ashok Kumar and Dr.I. Laurence Aroquiaraj

Diabetes mellitus (DM) commonly termed as diabetes is a physical condition of a human where the glucose level in the blood are very high. The classification of the Diabetic is very important and also a trivial task in-order to perform diagnosis and to interpret the diabetic data. The important issue in this regard is when the classification techniques are applied to these data sets, the performance is compromised. The objective of the research is to improve the level of accuracy in the predictive model and to bring out the model in such a way that it works well with other data sets also. In the latest work, K-Means algorithm is used but it suffers from the drawback of random initial values for the seed and also the overhead in setting the seed in lieu with the experience. This affects the performance in the classifier. To overcome this problem, an enhancement is made on the previous work by making the seed selection task of the samples of input using the Adaptive Divergence Weight Firefly Algorithm (ADWFA). The proposed research is aimed to develop an Adaptive Neuro Fuzzy Inference System (ANFIS) which is optimized and it is assumed that the values that are missing or in other words, the outliers are replaced with configurations using fuzzy logic which is expected to give better accuracy in high risk stratification. The proposed model is subjected to various pre-processing techniques and the model is primarily consisted of the Improved K-means algorithm with ADWFA (IKM-ADWFA) and the ANFIS algorithm. A new weight of inertia is used here and the same is calculated from the diverge function. The ADWFA achieves optimization and superiority in the task under consideration. The data set taken for the experiment is from The Pima Indians Diabetes and the experiment is carried out using MATLAB for analyzing the Knowledge and the results are compared with other outcomes using appropriate toolkits.

Volume 11 | 06-Special Issue

Pages: 18-31