Development of rough-TOPSIS algorithm as hybrid MCDM and its implementation to predict diabetes

Abstract

In this work, an innovative approach of multi-criteria decision-making method guided by rough set theory is researched to predict diabetes. Diabetes is the root cause of various deadly diseases. Designing an expert diabetes prediction model can solve the health monitoring issue with preventive measures beforehand. The proposed work has mainly two phases. In the first phase, the ensemble classification method develops the classification model, and rough set theory is implemented as a feature selection technique. In the second phase, TOPSIS, a multi-criteria decision-making method, is implemented for optimising classification models. Ensemble classification methods used here in this work: Bagging, AdaBoost, M1, Logit Boost, attributed selected classifier, random subspace, and multi-class classifier. The technique for order preference by similarity to ideal solution, the so-called TOPSIS, a multi-criteria decision-making method, has been used to select the optimised prediction model. Experimental diabetes data are collected from the UCI repository. Results obtained for predicting diabetes agree with those obtained from clinical practitioners.