Analysis on Diabetes Prediction Using Different Classifiers Techniques

Abstract

Huge amount of data is fetched from various resources. Data that is required from large data repository can be extracted with the use of data mining techniques, as there might be relevant and irrelevant data present in the repository. Thus minimizes the access of data from repository, so for increasing the data usage ability, data mining techniques are required. In data mining data is classified in different classes which are then used to implement a various disease prediction model. Similarly diabetes can be predicted using these prediction models based on different parameters. Parameters include Pregnancies, Blood Pressure, Skin Thickness, Glucose, Insulin, Diabetes Pedigree Function, Age and many more various classifiers. Different classifiers like CART, RF, SVM, LDA, KNN are being applied on PIMA dataset for the data at the confidence level of 0.95. CART has highest accuracy of 88%. Lowest accuracy is of KNN. So prediction using CART based classifiers is best.