Research on text classification of railway safety incidents based on BLS

doi:10.16265/j.cnki.issn1003-3033.2022.06.2262

Abstract

Abstract:

In order to prevent railway safety incidents, text mining related technologies and BLS were utilized to study effective incident classification mechanism, including four categories of equipment, construction, operation and external environmental problems. 314 pieces of text data were cleaned and structured, and Chinese word segmentation was built based on Jieba word segmentation + custom thesaurus+ custom stop word list. Then, 223 feature words were constructed based on Chi square test, and their weights were calculated based on TF-IDF. Finally, accident causes were classified according to BLS, and three classification methods were designed. The results show that the system can form an effective classification model through mining text information of railway safety event reports. And it can save computing power by utilizing features of BLS, and improve classification accuracy by adding feature enhancement nodes, so as improve industry management level.

Key words: broad learning system (BLS), railway safety incident, text classification, term frequency-inverse document frequency (TF-IDF), text mining

SHANG Linyu, YIN Ming, XIAO Chang, CHENG Jun. Research on text classification of railway safety incidents based on BLS[J]. China Safety Science Journal, 2022, 32(6): 103-108.

Figures/Tables 8

Fig.1

Tab.1

Fig.2

Tab.2

Fig.3

Tab.3

Tab.4

Fig.4

References 11

[1]	陈孝慈, 谭章禄, 单斐, 等. 基于Bigram的安全隐患文本分类研究[J]. 中国安全科学学报, 2017, 27(8):156-161. doi: 10.16265/j.cnki.issn1003-3033.2017.08.027
	CHEN Xiaoci, TAN Zhanglu, SHAN Fei, et al. Research on text categorization for hidden dangers based on Bigram[J]. China Safety Science Journal, 2017, 27(8):156-161. doi: 10.16265/j.cnki.issn1003-3033.2017.08.027
[2]	李浩然, 吴超仲, 褚端峰, 等. 基于K均值聚类的高速公路汽车碰撞事故分析[J]. 中国安全科学学报, 2013, 23(11):91-96.
	LI Haoran, WU Chaozhong, CHU Duanfeng, et al. Statistical analysis of car crash on freeways based on K-means clustering[J]. China Safety Science Journal, 2013, 23(11):91-96.
[3]	韩广, 卜桐, 王明明, 等. 基于双通道双向长短时记忆网络的铁路行车事故文本分类[J]. 铁道学报, 2021, 43(9):71-79.
	HAN Guang, BU Tong, WANG Mingming, et al. Text classification of railway traffic accidents based on dual-channel bidirectional long short tern memory network[J]. Journal of the China Railway Society, 2021, 43(9):71-79.
[4]	李新琴, 张鹏翔, 史天运, 等. 基于深度学习集成的高速铁路信号设备故障诊断方法[J]. 铁道学报, 2020, 42(12):97-105.
	LI Xinqin, ZHANG Pengxiang, SHI Tianyun, et al. Research on fault diagnosis method for high-speed railway signal equipment based on deep learning integration[J]. Journal of the China Railway Society, 2020, 42(12):97-105.
[5]	周庆华, 李晓丽. 基于MCNN的铁路信号设备故障短文本分类方法研究[J]. 铁道科学与工程学报, 2019, 16(11):2859-2865.
	ZHOU Qinghua, LI Xiaoli. Research on short text classification method of railway signal equipment faule based on MCNN[J]. Journal of Railway Science and Engineering, 2019, 16(11):2859-2865.
[6]	陈农田, 李俊辉, 满永政. 基于改进CNN-BiGRU-att模型的文本分类研究[J]. 昆明理工大学学报:自然科学版, 2022, 47(1):30-37.
	CHEN Nongtian, LI Junhui, MAN Yongzheng. Text classification based on improved CNN-BiGRU-att model[J]. Journal of Kunming University of Science and Technology:Natural Science, 2022, 47(1):30-37.
[7]	CHEN C L P, LIU Zhulin. Broad learning system: an effective and efficient incremental learning system without the need for deep architecture[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(1):10-24. doi: 10.1109/TNNLS.2017.2716952
[8]	SRIRAM B, FUHRY D, DEMIR E, et al. Short text classification in twitter to improve information filtering[C]. Research and Development in Information Retrieval,2010:841-842.
[9]	王杨, 许闪闪, 李昌, 等. 基于支持向量机的中文极短文本分类模型[J]. 计算机应用研究, 2020, 37(2):347-350.
	WANG Yang, XU Shanshan, LI Chang, et al. Classification model based on support vector machine for Chinese extremely short text[J]. Application Research of Computers, 2020, 37(2):347-350.
[10]	ANDRES A M, QUEVEDO M, GARCIA J M T, et al. On the validity condition of the chi-squared test in 2×2 tables[J]. Test, 2005, 14(1):99-128. doi: 10.1007/BF02595399
[11]	ZHAO Guifen, LIU Yanjun, ZHANG Wei, et al. TFIDF based feature words extraction and topic modeling for short text[C]. Management Engineering, Software Engineering and Service Sciences,2018:188-191.

分类编码	分类名称	细分类
F₁	设备问题	接触网断线、电缆故障、隧道问题、桥梁问题、断轨问题、钢轨损伤、信号设备故障、车载设备故障、供电设备故障
F₂	施工问题	路基下沉、路基上拱、损伤电缆、违章施工、施工不到位、施工侵限、碰触接触网
F₃	作业问题	司机违章开车、司机操纵不当、臆测行车、人员违规操作、调车问题、错误编组
F₄	外部环境问题	人为摆障、异物上道、进入闲杂人员、机动车肇事、治安问题、沿线火情、雨雪灾害

F₁		F₂		F₃		F₄
词条	p值	词条	p值	词条	p值	词条	p值
故障	6.22×10^-9	施工	3.82×10^-25	调车	6.29×10^-20	摆放	2.51×10^-6
分路不良	1.58×10^-6	挖断	6.37×10^-6	盲目	6.93×10^-14	地方	1.50×10^-5
占用	4.31×10^-5	拆除	3.52×10^-5	作业	1.86×10^-10	人为	7.47×10^-5
限速	5.67×10^-5	改造	0.000 10	挤坏	4.56×10^-8	闲杂人员	8.83×10^-5
热备	7.79×10^-5	红光带	0.000 14	关闭	1.27×10^-7	入网	0.000 21
内部	0.000 11	大修	0.000 22	越过	1.27×10^-7	防尘	0.000 21
牵引	0.000 41	大机	0.000 22	信号机	2.25×10^-6	防护网	0.000 10
交路	0.000 45	挖掘机	0.000 22	甩车	2.75×10^-6	碾轧	0.000 10
病害	0.000 77	电缆	0.000 52	值班员	8.06×10^-6	撞坏	0.001 25
更换	0.000 96	施工单位	0.000 71	抱闸	8.46×10^-6	落石	0.001 25

分类	训练数量	测试数量	合计
F₁	58	42	100
F₂	25	15	40
F₃	41	11	52
F₄	96	26	122
总计	220	94	314

模型	准确率/%	耗时/ms
BLS	87.23	3.9
BLS_Enhance	89.36	6.0
BLS_Feature&Enhance	91.49	5.0