基于随机森林的虚拟路谱准确度智能主观评价法

针对当前虚拟路谱准确度评价方法中的局限性,利用随机森林建立了一个虚拟路谱的智能主观评分模型,以路谱的客观统计值作为模型输入,以主观评价的分值作为输出. 在建模过程中全面考虑信号的多方面特征,利用特征重要度筛选出最重要的6个客观统计值作为模型输入; 通过模型训练固化多位专家的专业评分经验,避免每次进行主观评分时由人为因素导致的结果波动. 工程应用显示,该模型的精度高、泛化能力强,是对虚拟路谱评价的一个全面、高效、准确的智能工具.

Intelligent Subjective Evaluation of Virtual Road Load Data Accuracy Based on Random Forest

Considering the limitations of current methods for virtual road load data evaluation,an intelligent subjective evaluation model of virtual road load data signals based on random forest is proposed. The objective statistical parameters of time history signals are taken as the inputs and the scores of subjective evaluations from experts as the output. The most important six objective statistical parameters are selected through feature importance ranking,which takes various features of the time history signals into a comprehensive consideration. The professional evaluation experiences from many experts are solidified into the model by training,and the fluctuations caused by human factors is avoided. The intelligent subjective evaluation model has high accuracy and good generalization,and it is proven a comprehensive,efficient and accurate intelligent tool for virtual time history signal evaluation for real-world applications.