Article Preview
TopIntroduction
Diabetes is regarded as a very baleful and constantly recurring disease among a set of health-related problems. That is also referred to as 'Diabetes Mellitus, and it is considered one of the foremost reasons for deaths in India. The disease is constantly recurring and occurs when the pancreas does not create sufficient insulin levels. (World Health Organization, 2003) It is also happening in the situation when the affected person is incapable of utilizing the produced insulin. The regulatory function of insulin is to maintain a suitable blood sugar level. The increased sugar level in a person's blood is generally visible and could affect the nervous system and blood cells. Diabetes can also cause other diseases such as blindness, kidney failure, jolting, and cardiovascular disease. Recent research has revealed that approximately 98 million persons might be affected by diabetes in India by 2030 (Weir. & Bonner-Weir, 2004). Hence, there is a need to diagnose and prevent the disease at an early stage.
Diabetes is categorized as Type-1, Type-2, and Gestational diabetes
Type-1 of the disease (diabetes): Type-1 or 'Juvenile category of diabetes' is found when the person's body is incapable of generating a proper insulin level. Since the person suffering from this disease category has a dependency on insulin, it is advised that the person have an insulin intake, which is artificially available (Lee et al., 2011).
Type-2 of the disease (diabetes): Type 2 and impacts insulin use by the body cells. Although the generation of insulin is satisfactory, the cells cannot respond and adequately use insulin. Most diabetes cases are of the type-2 category (Jayanthi & Babu, 2017).
Gestational category of the disease (diabetes): This disease category is found in females during pregnancy as the body reflects a lower degree of sensitivity to insulin. This disease category is not found in every female and is generally resolved after the delivery.
In this paper, the authors propose a machine learning-based scheme to analyze sample data sets of PIMA to classify the data and forecast the presence or absence of diabetes (Soofi, 2017; Patel, 2017)). The emphasis is to assess the outputs of classification-based frameworks and forecast the chances of occurrence of diabetes in a person with the highest level of accuracy (WHO, 2011; WHO, 2016). Here authors have implemented nine different classification models: AdaBoost Classifier, Logistic Regression, GaussianNB, Bagging Classifier, Decision Tree Classifier, k-NN Classifier, Voting Classifier, Random Forest Classifier, Gradient Boosting Classifier to perfectly examine the dataset (Greenwood et al., 2015).
The architecture of the proposed approach is shown in Figure-1 entitled 'Diabetic data pre-processing'.
Figure 1. Diabetic data pre-processing
TopBackground
Although numerous machine-learning algorithms are used in this research, several representations and forms capable of forecasting a category of diabetes are evaluated for their exactness. The selected algorithms with the highest level of accuracy are discussed, and their results are compared (Perveen et al., 2019). Currently, designing monitoring models for actual data of patients is getting attraction, and many models are available for use in the health monitoring system (Chauhan et al., 2019).