Diabetes Mellitus prediction using Improved Chaotic Whale Optimization and Data mining techniques

Abstract

Automation of disease detection is now prevalent in healthcare systems. Diabetes mellitus is a serious issue that has become prominent worldwide. It is a hereditary illness that impairs human existence at every stage of development. Each year, millions more people develop diabetes, which also has an impact on young people. Currently, the medical industry is shifting toward automation, as the process of identifying diseases needs periodic human verification. A single model for diabetes prediction is insufficient for complicated problems since it might not be appropriate for training huge data. In this proposed approach, a novel method which combines the outcomes of more than one algorithm is suggested. An ensemble approach using data mining techniques such as Decision Trees, Random Forests, AdaBoost, Gradient Boosting and XGBoost is employed to predict diabetes mellitus. Feature selection methods such as Averaged Fisher score and Kolmogorov-Smirnov score are used to choose the prominent characteristics from PIMA Indian Diabetes dataset. Further, the most optimal features are selected using Improved Chaotic Whale Optimization algorithm. The proposed approach produces predictions that are 98.8% accurate and reliable.