Logistic regression and Random forest-based hybrid classifier with recursive feature elimination technique for diabetes classification

Abstract

Diabetes mellitus is a chronic metabolic ailment being considered as one of the deadliest diseases in the world. Millions of cases and deaths have been reported due to diabetes. Initial forecast of diabetes condition helps in reducing the death rate significantly that happen due to it. Current advancements in biomedical techniques have facilitated to store the electronic health record datasets which can be analyzed for better diagnosis. In order to explore these datasets, data mining techniques are considered as promising techniques which examines the data rapidly and provides a desired outcome. In this work, the data mining technique was adopted for diabetes classification using machine learning techniques. The proposed approach comprises of several steps such as missing value imputation, attribute selection and classification which are performed using mean missing value imputation, logistic regression &Recursive Feature Elimination for feature selection and random forest for classification, respectively. The evaluation results demonstrate that projected methods achieve97.39% prediction accuracy which shows a significant improvement in contrast to prevailing current approaches.