A Hybrid Feature Selection Mechanism

This paper uses the SVM to predict the protein disordered region. Nevertheless, the number of features used in this paper is 440. Both time and space complexity is high while performing the support vector machine (SVM) training and testing. So this paper proposes a hybrid feature selection mechanism to reduce the dimensionality of the feature space. The filter and wrapper feature selection methods are combined to improve the SVM´s predictability and decrease the processing time. First, two filters are used to screen out redundant features. The resulted feature subsets are then combined for the wrapper method to do final fine tuning. The results demonstrate the usefulness of this hybrid mechanism.