基于随机森林结合前向序列选择算法的铁路隧道防水板检测数据关键影响因素识别

针对参数特征复杂度高的铁路产品,提出了一种基于随机森林(RF)结合前向序列选择(SFS)算法的铁路产品检验检测数据关键影响因素识别方法,以辅助基于经验的识别方法。创新使用RF-SFS算法,将其应用于铁路隧道防水板检验检测数据关键影响因素的识别研究。根据多年铁路隧道防水板检测数据,建立RF模型,获得了影响铁路隧道防水板检测结果的特征关键性评分序列。随后,结合SFS方法得出关键性评分序列的阈值,将排名前6位的影响因素识别为关键特征,模型的预测能力达到99.98%。为验证关键特征识别方法的有效性,对比分析3种模型在使用不同特征子集时的预测能力。当仅选用关键特征时,3种模型的预测能力均达到最佳,加入冗余特征后模型的预测能力逐渐降低。

Key Influencing Factors Identification of Railway Tunnel Waterproof Board Detection Data Based on RF-SFS Algorithm

Aiming at railway products with high complexity of parameter features,a method for identifying key influencing factors of railway product testing based on Random Forest(RF)and Sequential Forward Selection(SFS)algorithm was proposed to assist the experience-based identification method.The RF-SFS algorithm was innovatively used and applied to the research on the identification of the key influencing factors of the railway tunnel waterproof board detection data.Based on the detection data of railway tunnel water proof board for several years,a RF model was established,and the characteristic key score sequence of the factors affecting the detection result of the railway tunnel waterproof board was obtained.Subsequently,combined with the SFS method,the threshold of the key score sequence was obtained,and the top 6 influencing factors were identified as key features.The prediction ability of the model reached 99.98%.In order to verify the effectiveness of the method,the predictive ability of the three models when using different feature subsets was compared and analyzed.When only key features are selected,the predictive capabilities of the three models are all the best.After adding redundant features,the predictive capabilities of the model gradually decrease.