Proposed Hybrid Attribute Selection Method on Financial Data Sets

One of the most important problems encountered in data analysis is the removal of variables in the data that create noise and affect the solution negatively. The most important method used to solve this problem is feature selection. In the paper, a hybrid method is proposed for feature selection on financial data. In the literature, feature selection is examined in 3 categories: filtering, wrapper and recursive methods. In the proposed hybrid method, 2 filtering, 2 buried and 1 spiral method are utilized. As a result of the studies on 2 different financial data, the proposed method successes as good results as the best methods in the literature. The study showed no single feature selection method to use for each data set. In addition, scaling in accordance with the data increased the success rate.