基于随机森林算法的地铁通信系统健康度动态评估方法

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

[目的]地铁通信系统的健康度直接关系到地铁运营的效率和安全。针对现阶段地铁通信系统健康度评估方式存在评估结果片面性及评分标准主观性等问题,需构建能够科学、准确地评估地铁通信系统运行状态和健康度的方法。[方法]提出了一种基于随机森林算法的健康度动态评估方法,结合深度学习进行通信健康度检测。设计评价指标体系,收集相应数据,在标签(真实健康度)不明确的情况下,采用混合标签法确定标签并计算各个指标重要度;适当地加入专家知识的影响,根据指标权重和专家制定的指标自身评分标准,计算通信系统各子系统健康度;基于成都地铁维保数据进行了该方法试验验证。[结果及结论]试验验证结果表明,通过该方法可以得到有效的指标重要度,可合理评估地铁通信系统健康度。

[Objective]The health of metro communication system has a direct bearing on the efficiency and safety of met-ro 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 nec-essary to construct a method that can scientifically and accu-rately 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 de-signed to collect the corresponding data.The mixed label meth-od 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 incor-porates the influence of experts knowledge.Based on the ma-intenance 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.