Main Article Content
People's busy lifestyle and food habits throw them into the pool of many lifestyle diseases; diabetes is one of those. Overall, 422 million people live with diabetes as per World Health Organization (WHO) 2014 statistics. This is particularly severe in the middle- and low-income countries irrespective of gender and age. Much research has been done in predicting diabetes disease but primarily focused on symptoms and diagnostic test reports. Although there are many existing algorithms available, no single classifier is providing an optimal solution. Therefore, in this work, we primarily focus on diabetes occurrence risk prediction by ensembling various machine learning techniques. We also aimed at predicting the risk of diabetes among men and women based on various symptoms such as Polyuria, Polydipsia, Sudden weight loss, Weakness, Polyphagia, Genital thrush, Visual blurring, Itching, and Irritability. Various combinations of algorithms are taken in an ensemble approach during experimentation. Finally, the combination of Naïve Bayes Classifier, SVM Classifier, J48, Optimized Parametric Multilayer Perceptron are performing better when compared with other combinations. All the experiments were carried out on Kaggle data science datasets. The experiments show that the ensemble classifier predicts the disease with almost 100% accuracy on the benchmark dataset. Further, the experimentation can be extended to predict the level of occurrence and the occurrence of other health complications because of this disease.