Decision tree classification algorithm based on cost and benefit dual-sensitive

Decision tree classifier based on cost-sensitive is a hot research direction in recent years. Although this method can get decision results with lower cost, there are some limitations in practical application for default considering the benefits of correct classification. This article defines the conception of correct classification benefit, and then builds a novel decision tree based on cost and benefit dual-sensitive (CBDSDT). In order to obtain the best classification result with lower cost and higher benefit, our method takes into account the test cost, misclassification cost, attribute information and correct classification benefit. Experiments demonstrate that our method has better usability and stability.