Analytical model to predict diabetic patients using an optimized hybrid classifier.

Diabetes is the most common disease and is a major cause for blindness, kidney failure, heart attacks, stroke and lower limb amputation. Thus, early prediction of diabetes is very crucial to initiating proper treatment to avoid further serious complications of the disease. The performance of recent diabetes detection schemes based on clinical data is highly influenced by low feature distinctiveness and unwanted features such as dermatologic manifestations. Different machine learning classifiers need tedious hyper-parameter tuning, which fails to assure a better diabetes detection rate. This article presents an analytical model to detect diabetes based on an optimized Support Vector Machine (SVM), K Nearest Neighbor (KNN) and Random Forest (RF) using decision level fusion to improve the diabetes detection rate. The hyper-parameters SVM, KNN, and RF are optimized using a multi-objective function-based Particle Swarm Optimization (PSO) algorithm, which considers various clinical entities for the diabetes detection, such as age, body mass index (BMI), blood pressure (BP), glucose, insulin, number of pregnancies, skin thickness, and diabetes pedigree function. The extensive experiments on the Indian Pima diabetes dataset confirmed that the diabetes detection using hybrid classifiers can provide a better prediction rate (94.27%) compared with single classifiers and the previous state of arts. [ABSTRACT FROM AUTHOR]