Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
10360416 | Pattern Recognition | 2014 | 10 Pages |
Abstract
According to our experimental evaluation using UCI-repository datasets, LCMmax.R was about 5-11 times faster than LCMmax.Rnaive, which indicates effectiveness of the introduced two improvements. MRF repetition, however, was significantly faster than LCMmax.R, and it was fast enough for practical usage. The experimental results using UCI-repository datasets also showed that accuracy of a nearest rectangle classifier using an RGC is close to that using the hyperrectangles output by the randomized subclass method (RSM) [2] though the number of component rectangles of an RGC is significantly smaller than the number of the hyperrectangles output by RSM. The performance of RGC was also shown to be comparable to that of the six popular classifiers including logistic regression and support vector machine. The disjunctive normal form representation of the classification rules obtained by RGC was demonstrated to be simpler and more readable for us than that obtained by RSM and C4.5.
Keywords
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Vision and Pattern Recognition
Authors
Koji Ouchi, Atsuyoshi Nakamura, Mineichi Kudo,