A multiple kernel support vector machine scheme for simultaneous feature selection and rule-based classification

Authors:
Zhenyu Chen;Jianping Li
Affiliations:
Institute of Policy & Management, Chinese Academy of Sciences, Beijing, China and Graduate University of Chinese Academy of Sciences, Beijing, China;Institute of Policy & Management, Chinese Academy of Sciences, Beijing, China
Venue:
PAKDD'07 Proceedings of the 11th Pacific-Asia conference on Advances in knowledge discovery and data mining
Year:
2007

Citing 5
Cited 0

Choosing Multiple Parameters for Support Vector Machines

Machine Learning
Symbolic Interpretation of Artificial Neural Networks

IEEE Transactions on Knowledge and Data Engineering
Rule extraction from linear support vector machines

Proceedings of the eleventh ACM SIGKDD international conference on Knowledge discovery in data mining
Learning the Kernel Function via Regularization

The Journal of Machine Learning Research
Feature subset selection for support vector machines through discriminative function pruning analysis

IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

Quantified Score

Hi-index	0.00

Visualization

Abstract

In many applications such as bioinformatics and medical decision-making, the interpretability is important to make the model acceptable to the user and help the expert discover the novel and perhaps valuable knowledge hidden behind the data. This paper presents a novel feature selection and rule extraction method which is based on multiple kernel support vector machine (MK-SVM). This method has two outstanding properties. Firstly, the multiple kernels are described as the convex combination of the single feature basic kernels. It makes the feature selection problem in the context of SVM transformed into an ordinary multiple parameters learning problem. A 1-norm based linear programming is proposed to carry out the optimization of those parameters. Secondly, the rules are obtained in an easy way: only the support vectors necessary. It is demonstrated in theory that every support vector obtained by this method is just the vertex of the hypercube. Then a tree-like algorithm is proposed to extract the if-then rules. Three UCI datasets are used to demonstrate the effectiveness and efficiency of this approach.