Abstract:

Abstract In ancient times, an accurate diabetes prediction and type of classification are the most important and demanding tasks in the medical field for providing proper diagnosis to the patients. For this purpose, various machine learning based detection systems are developed in the conventional works to predict the diabetes from the given dataset. Still, it has some limitations with the factors of difficult to understand, high time requirement for training and testing, over fitting, and error outputs. Therefore, the proposed research work objects to implement a group of data mining techniques for developing an automated and efficient diabetes detection system.  In this framework, an Inherent Coefficient Normalization (ICN) technique is implemented at first for preprocessing the PIMA Indian dataset obtained from the repository, which highly improves the quality of data for processing. Then, an Intelligent Harris Hawks Optimization (IHHO) technique is utilized to optimally select the features for training the classifier. Finally, the Pivotal Decision Tree (PDT) based classification technique is deployed to predict the data as whether diabetes or non-diabetes with reduced computational complexity and time consumption. During analysis, the performance and results of the proposed IHHO-PDT technique is validated and compared using various measures.

Keywords: