Further Experiments on A Combination of Linear SVM Weight and ReliefF for Dimensionality Reduction

This research further investigated how dimensional data could be efficiently downsized using a multilayered technique based on a combination of two major feature selections, including Linear SVM Weight and ReliefF together with classifier namely Support Vector Machine (SVM). Two datasets, including SRBCT and USPS, were used for the experiment. The results show that the proposed technique is more efficient than using either Linear SVM Weight or ReliefF alone for dimensionality reduction. The dimensional data could be downsized from 2,308 to 8 attributes where the accuracy rate could reach 100 percent in SRBCT. The experimental result of SBRCT was also consistent with that of USPS in which the dimensional data could be downsized from 256 to 55 attributes with the accuracy of 95.76 percent.