Name
Abstract | ||
|---|---|---|
Support vector machines (SVMs) are learning systems based on statistical learning theory that have been applied with excellent generalization performance to a variety of applications in classification and regression. However, as Artificial Neural Networks, SVM are black box models, that is, they do not explain the process by which a given result is attained. Some models that extract rules from trained SVM have already been proposed but the rules extracted from these methods use, in the antecedents, crisp intervals or functions, which greatly reduce the rule's interpretability. Therefore, a fuzzy rule extraction method from trained SVM, called FREx_SVM, was previously developed, providing rules with good accuracy and coverage of the database. However, the classification performance of the extracted rules was usually much lower than the original SVMs. To improve this performance we have developed an extension of FREx-SVM, where the fuzzy sets are automatically tuned so to attain better classification performance without reducing the interpretability of the extracted fuzzy rules. |
Year | DOI | Venue |
|---|---|---|
2013 | 10.1109/IFSA-NAFIPS.2013.6608427 | PROCEEDINGS OF THE 2013 JOINT IFSA WORLD CONGRESS AND NAFIPS ANNUAL MEETING (IFSA/NAFIPS) |
Keywords | Field | DocType |
svm,support vector machines,accuracy,neural nets,genetic algorithms,artificial neural networks,fuzzy set theory,fuzzy sets,kernel,statistical analysis,optimization | Neuro-fuzzy,Fuzzy classification,Defuzzification,Least squares support vector machine,Pattern recognition,Fuzzy set operations,Support vector machine,Fuzzy set,Artificial intelligence,Fuzzy number,Machine learning,Mathematics | Conference |
Citations | PageRank | References |
0 | 0.34 | 7 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| C. F. F. Carraro | 1 | 0 | 0.34 |
| Marley B. R. Vellasco | 2 | 280 | 47.47 |
| Ricardo Tanscheit | 3 | 118 | 21.53 |