Discretization based Support Vector Machines for Classification of Spatial Data.

Abstract

Discrete values have important roles in data mining and knowledge discovery. They are about intervals of numbers which are more concise to represent and specify, easier to use and comprehend as they are closer to the knowledge level representation than continuous ones. Discretization is the process of quantizing continuous attributes. It has been used for decision tree classifier. The success of discretization can significantly extend the borders of many learning algorithms. Support Vector Machines (SVM) are the new generation learning system based on the latest advances in statistical learning theory. SVM is the recent addition to the toolbox of data mining practitioners and are gaining popularity due to many attractive features, and promising empirical performance. In this paper, a new approach to classify data using SVM classifier, after discretization is looked into. The classification results achieved after discretization based SVM are much better than the classification results using simple SVM in terms of accuracy. To acquire the better accuracy, discretization has been instrumental This is an attempt to extend the boundaries of discretization and to evaluate its effect on other machine learning techniques for classification namely, support vector machines.