General Approaches For Feature Selection

1468 WordsJun 7, 20166 Pages
General Approaches for Feature Selection There are 3 types of approaches for feature selection namely filter, wrapper, embedded method. Filter method: Filter method does not involve a learning algorithm for measuring feature subset [6]. It is fast and efficient for computation .filter method can fail to select the feature that are not beneficial by themselves but can be very beneficial when unite with others. Filter method evaluates the feature by giving ranks to their evaluation value. In filter method it evaluates the correlation between the features by using criteria such as, mutual information, maximum length, maximum relevance min redundancy (mRMR), PCA. Figure 1.2 Filter approach Wrapper method: wrapper method involve learning algorithm and search for optimal feature subset from original feature set which discover the relationship between relevance and optimal subset selection. It performed better than the filter method. The specific training classifier is used to evaluate the performance of selected features. Figure 1.3 Wrapper approach Embedded method: Embedded method is a combination of wrapper method and embedded method. This decreases the computational cost than wrapper approach and captures feature dependencies. It searches locally for features that allow better discrimination and also the relationship between the input feature and the targeted feature. It involves the learning algorithm which is used to select optimal subset among
Open Document