THORS: An Efficient Approach for Making Classifiers Cost-Sensitive

In this paper, we propose an effective THresholding method based on the Order Statistic, called THORS, to convert an arbitrary scoring-type classifier, which can induce a continuous cumulative distribution function of the score, into a cost-sensitive one. The procedure uses the order statistic to find an optimal threshold for classification, requiring almost no knowledge of the classifier itself. Unlike common data-driven methods, we analytically show that the THORS has theoretical guaranteed performance, theoretical bounds for the costs, and low-time complexity. Coupled with empirical results on several real-world data sets, we argue that the THORS is the preferred cost-sensitive learning technique.