Statistical Measures to Determine Optimal Structure of Decision Tree: One versus One Support Vector Machine
In this paper, one versus one optimal decision tree support vector machine (OvO-ODT SVM) framework is proposed to solve multi-class problems where the optimal structure of decision tree is determined using statistical measures, i.e., information gain, gini index, and chi-square. The performance of proposed OvO-ODT SVM is evaluated in terms of classification accuracy and computation time. It is also shown that proposed OvO-ODT SVM using all the three measures is more efficient in terms of time complexity for both training and testing phases in comparison to conventional OvO and support vector machine binary decision tree (SVMBDT). Experiments on University of California, Irvine (UCI) repository dataset illustrates that ten crossvalidation accuracy of our proposed framework is comparable or better in comparison to conventional OvO and SVM-BDT for most of the datasets. However, the proposed framework outperforms the conventional OvO and SVM-BDT for all the datasets in terms of both training and testing time.
Defence Science Journal, 2010, 60(4), pp.399-404, DOI:http://dx.doi.org/10.14429/dsj.60.500
Where otherwise noted, the Articles on this site are licensed under Creative Commons License: CC Attribution-Noncommercial-No Derivative Works 2.5 India