A cooperative GA-SM based prediction model for healthcare services

Abstract

Diabetes mellitus is a major health challenge around the world. The blood glucose level is one of the major factors in the human body and a significant increase in its level can cause many harmful effects in human life. It is expected that early diagnosis of diabetes mellitus can lead to rapid and effective treatment of glycemic control. As the number of people who suffer from diabetes mellitus increases significantly, a study on diabetes mellitus prediction was done through well-known methods in data mining (DM). In this paper, a genetic algorithm (GA)-based suppressor mutation (SM) optimisation rule miner has been proposed as a cooperative approach for prediction of diabetes mellitus. A novel fitness function has been incorporated into the GA-SM approach to generate a comprehensive optimal rule set while balancing accuracy, sensitivity and specificity. The proposed rule miner was compared against three rule-based algorithms, namely CN2, J48 and BF tree on the Pima Indians Diabetes Dataset with 768 patient records using ten-fold cross validation. The results obtained prove that the proposed GA-SM approach has outperformed CN2, J48 and BF tree with respect to accuracy and kappa.