Optimal Feature Subset Selection with Errors and Variable Costs
Liao Shu-jia · 2015
Cost is important to data in real application. Test costs of data are sensitive to error ranges, namely, the granularity of data, while misclassification costs are also related to test costs. The existing feature selection methods often ignore it. To address this situation, an approach is proposed for selecting optimal feature subset with error ranges and variable costs. Firstly, the theoretical framework is established. Then the corresponding algorithms are designed. In the method, test costs and misclassification costs are adaptively computed according to the confidence level of measurement errors. The objective of feature selection is to minimize the average total cost. By the method, the optimal feature subset and the best confidence level of errors can be obtained. The experimental results manifest the effectiveness of the proposed approach.