Authors : Wipawan Buathong and Phayung Meesad
Abstract: This study proposes an alternative feature selection technique for dimensionality reduction namely Double Linear Support Vector Machine or “DLSVM” Weight. The efficiency of DLSVM was measured based on four performance evaluation criteria (i.e., accuracy, F-measure, precision and recall). The efficiency of well recognised feature selection techniques was also measured for comparative purposes. The Support Vector Machine (SVM), a prominent classifier was also used with DLSVM and these feature selection techniques. The Leukemia dataset from the University of California Irvine (UCI) machine learning repository was used for the experiments. Downsized data dimensions were classified into 60, 50, 40, 30, 20 and 10, respectively. The experimental results showed that the DLSVM was much more efficient than other feature selection techniques at almost all of the data dimensions. Particularly, all performance evaluation criteria of DLSVM could reach 100% when original data dimensions were downsized from 5,147-60, 50 and 40.
Wipawan Buathong and Phayung Meesad, 2014. Double Linear Support Vector Machine for Dimensionality Reduction. Research Journal of Applied Sciences, 9: 208-213.