The state or disorder where the body cannot effectively use the insulin is called Diabetes. If the insulin levels are not maintained properly, the diabetes is one such disorder where it damages all other body parts. It is estimated that the diabetes is the 7th leading cause of deaths as per World Health Organisation report. Early recognition of diabetes, decreases the risk of serious ailments, which includes, heart diseases, brain stroke, eye related diseases, kidney diseases, nerve related diseases etc. In the present work, pima indians diabetes data set is considered as the best dataset and different models viz., hierarchical clustering with decision tree, hierarchical clustering with support vector machines, hierarchical clustering with logistic regression and k means with logistic regression are developed and implemented for identifying and predicting the diabetes. The accuracies of these prediction models range between 0.90 and 0.946. An Improved Diabetes Prediction Algorithm (IDPA) combining the hierarchical clustering algorithm and Naïve Bayes classification algorithm is developed to identify and predict the Type-II diabetes and has shown an accuracy of 0.96. In this IDPA, firstly, the grouping of data into two groups i.e. diabetes and non-diabetes is done by applying the hierarchical clustering algorithm. Then, the filtering is done by comparing the group value to the class value followed by applying Naïve Bayes classification algorithm for predicting diabetes. The results show that the proposed novel method i.e. IDPA can predict the diabetes with higher accuracy levels (0.96) than the traditional/existing methods and other methods which were implemented. This model can be used to predict diabetes early, thereby reducing the serious complications of diabetes.
Improved logistic regression model for diabetes prediction by integrating PCA and K-means techniques
