International Journal of Pharma and Bio Sciences
ijpbs.net
editorijpbs@rediffmail.com (or) editorofijpbs@yahoo.com (or) prasmol@rediffmail.com
10.22376/ijpbs.2019.10.1.p1-12
Volume 5 Issue 3
2014 (July- September)
PERFORMANCE EVALUATION OF SUPPORT VECTOR MACHINE – NEAREST NEIGHBOR CLASSIFIER FOR DIABETES DATASET
In this current scenario the automatic classification has been used for many applications such as indexing for information retrieval system, search engines, document organization and text filtering. Classification is one of the widely used techniques in the machine learning. It is a mechanism of grouping the data according to the predefined class labels. The popular classification algorithms are K-Nearest Neighbor algorithm (K-NN) and Support Vector Machine (SVM) algorithm. K-NN is a lazy learning method where classification is done by comparing the feature vectors of different points. K-NN is popular due to its simplicity and efficiency, but complexity is in finding the k value as a small k value will result in obtaining lesser information from training data. Support Vector Machine is another algorithm where an optimal hyperplane is chosen for classifying the diabetes dataset. SVM provides high accuracy and found to be popular in high dimensional data space. The hybrid classification algorithm is used to build the classification model using SVM-NN classifier is proposed. In this proposed SVM-NN classifier the impact of k parameter is reduced by considering only support vectors in order to classify the data. In SVM-NN Manhattan distance measure is used to compute the distance between the test samples and support vectors. The test samples can be compared to the class labels of the original class labels and performance can be evaluated using the confusion matrix. This proposed SVM-NN algorithm can reduce the size of the training samples and also greatly reduces the classifying time, so it can be used for large data sets. The experimental result shows that SVM-NN gives the best performance for Diabetes dataset.
D. MAHALAKSHMI AND S.P. CHOKKALINGAM
Classification, K-Nearest Neighbor, Support Vector Machine, Manhattan Distance Measure.
728-740