Dynamic Health Evaluation Method for Metro Communication System Based on Random Forests Algorithm

Abstract

Objective The health of metro communication system has a direct bearing on the efficiency and safety of metro operation. In view of the problems such as one-sided evaluation results and subjective evaluation criteria in current health evaluation method for metro communication systems, it is necessary to construct a method that can scientifically and accurately evaluate the operation status and health of the metro communication system. Method A dynamic health evaluation method based on random forests algorithm is proposed, and a communication health inspection is conducted in combination with deep learning. The evaluation index system is designed to collect the corresponding data. The mixed label method is used to determine the label and calculate the importance of each index when the label (true health degree) is not clear. The health degree of each subsystem of communication system is calculated according to the index weight and the index self scoring standards established by experts, which properly incorporates the influence of experts' knowledge. Based on the maintenance data of Chengdu Metro, the method is tested and verified. Result & Conclusion The test verification results show that with the above method, effective index importance can be obtained and the health of metro communication system can be reasonably evaluated.

Keywords

• metro
• communication system
• health evaluation
• random forests algorithm