Classification of Pima Indian Diabetes Dataset Using Support Vector Machine with Polynomial Kernel

A large population of the world is suffering from diabetes, which adversely affects the entire human body. If this type of disease is not diagnosed at an early stage it may cause serious health issues at a later stage. So an early detection of diabetes is very important to minimize health risks. Diabetes can be preventable, and if proper measures are taken at an early stage it can lower the chances of heart disease, cancer, stroke, kidney, and nerve damage. Supervised machine learning algorithms are more accurately used for classification challenges. Supervised machine learning algorithms discover a relationship between the input variables and target variables. The relationship is stored in a structure called model. These models are used for classification purposes. This chapter aims at identification of diabetes using support vector machine (SVM) in Pima Indian heritage. SVM can solve both linear and nonlinear problems. SVM uses mathematical functions called kernels to transfer original data and creates a hyperplane based on the transformation. The hyperplane created separates the data into different categories using support vectors. In this chapter SVM with polynomial kernel is proposed, with a different degree to attain a high level of accuracy. The learning algorithm is applied on the Pima Indian Diabetes dataset. The outcomes of the proposed model are compared with other models like SVM with radial basis function (RBF) kernel, SVM with sigmoid kernel, SVM with linear kernel, and data mining classifiers like artificial neural network (ANN) and C5.0 Decision trees (DT). The performance of the algorithm is also measured using the statistical indicators: accuracy, specificity and sensitivity. Gain graph and response graph are also used for evaluation of the performance of the model. The experimental study reveals that the outcomes of the proposed model are superior as compared to other models. This model can be effectively used as a decision support system in healthcare for identification of diabetes at an early stage.