Text-mining-based Automatic Classification Of Environmental Hidden Dangers Outside Railways

With the continuous expansion of the railway network in China,the external environmental problems of the railway become more and more prominent,which has become one of the important risk sources of railway traffic safety.For this reason,the party and state leaders pay great attention to this.In 2019 and 2020,the party and state leaders made important instructions and instructions for the governance of railway external environmental hazards for many times.China National Railway Group Co.,Ltd.(hereinafter referred to as "China Railway Group")attaches great importance to the safety of the external environment of the railway,and has carried out many special activities to investigate and rectify the hidden dangers of the external environment of the railway.In October 2019,the Safety Supervision Bureau of China Railway Group organized a nationwide investigation and remediation of high-speed rail external environmental hazards.Due to the different management characteristics of each Road Bureau and the different understanding of the management methods of external environmental hazards,the problems that the external environmental hazards should belong to(at present,China Railway Group divides the external environmental hazards into 15 categories and 76 categories)are not clearly identified in the investigation and remediation actions.This has seriously affected the State Railway Group and the railway bureaus to accurately grasp the situation of external environmental hazards and implement the main responsibility of the local government.Based on text mining and support vector machine(SVM),this paper realizes the automatic classification of external environmental hazards.This paper mainly completes the following work:(1)Build a thesaurus of hidden dangers in the external environment.Firstly,the description text of external environmental hazards is cut into words or single word sets by Chinese word segmentation,de stop words and other means to build the external environmental hazards feature thesaurus;secondly,the word and word weight in the feature thesaurus are optimized by Chi test and TF-IDF algorithm;finally,based on the optimized feature thesaurus,the external environmental hazards are realized by text mining technology Describes the vectorization of text.(2)Based on the vector description text of external environmental hazards,SVM classification algorithm is used to realize the automatic classification of external environmental hazards.First,based on the SVM classification algorithm and combining the characteristics of hidden data in the external environment of the railway,two classifiers,EHD-CSVC and EHD-VSVC,are constructed.Based on the optimized thesaurus,the classification effects of the two classifiers under the conditions of the polynomial,RBF,and sigmoid kernel functions are compared.The results show that the EHD-CSVC classifier under the polynomial kernel function has the best classification effect.Under the condition of selecting kernel function and classifier,the kernel function and classifier parameters are optimized.When the polynomial order is 3,the penalty coefficient is 32,and the kernel function parameter γ is 3.0517578125e-5,the classification accuracy can reach 92.33%.(3)The ehd-csvc classifier is implemented systematically to meet the actual work requirements.Based on the above research results,relying on the Railway External Environmental Safety Management Information System,the function of automatic identification of the external environmental hazards of hidden dangers is realized,and the function test is carried out with some external environmental hazards data that are not involved in machine learning.The classification accuracy is 90.5%,and the average response time is 1.04 s.The test results show that the accuracy and response speed of the automatic hidden danger classification function can meet the needs of daily work.Finally,the hidden danger data is analyzed based on the function of automatic identification of hidden danger category of external environment.Figure 30,table 9,reference 49.