The Fault Diagnosis Model for Railway System Based on an Improved Feature Selection Method

Finding reasons for the railway system faults is an important task for guaranteeing the safety of railway system. However, it usually takes a lot of time for people to do that task since the text data written by natural language to describe the faults are often very large. To overcome that problem by using computers, a diagnosis model can be built to process and classify those text data so as to find the causes of the railway system faults. But how to improve the accuracy of the diagnosis model is also an important issue. In this paper, we build a fault diagnosis model for railway system and improve the accuracy of the model by proposing a MI-RFE (Mutual Information and Recursive Feature Elimination) feature selection method. Compared with the traditional MI (Mutual Information) feature selection method, the MI-RFE feature selection method proposed in this paper improves the accuracy of diagnosis model of the railway system fault by 9%.