交通事件网络舆情分析方法  

An Analysis Method of Online Public Opinions on Traffic Incidents

作　　者：滕靖[1,2] 刘韶杰 龚越 王文 TENG Jing;LIU Shaojie;GONG Yue;WANG Wen(The Key Laboratory of Road and Traffic Engineering of Ministry of Education,Tongji University,Shanghai 201804,China;Shanghai Key Laboratory of Rail Infrastructure Durability and System Safety,Tongji University,Shanghai 201804,China;Hangzhou Hikvision Digital Technology Co,Ltd,Hangzhou 310052,China)

机构地区：[1]同济大学道路与交通工程教育部重点实验室,上海201804 [2]同济大学交通运输工程学院,上海201804 [3]杭州海康威视数字技术股份有限公司,杭州310052

出　　处：《交通信息与安全》2019年第6期139-148,共10页Journal of Transport Information and Safety

基　　金：国家社会科学基金项目（18BGL005）资助

摘　　要：交通运输行业关系国计民生,感知民意是政府做好交通治理的必要条件。在互联网时代,通过网络获取并分析交通事件舆情具有样本丰富、信息客观、获取及时等特点。基于微博、微信、新闻客户端等电子信息来源,提取交通事件舆情特征,形成交通事件的网络舆情系统分析方法。包括:获取网络舆情数据并进行文本预处理,构造交通事件舆情主题分类模型,建立评价重要度计算方法,结合舆情生命周期分析舆情演变趋势,建立交通舆情情感库并结合机器学习分析情感状态演化,通过关键词可视化定位舆情事件要点,实现网络舆情信息与可视化分析技术的耦合。该方法文本分类评估结果F值高于80%,情感极性判断准确率高于通用SnowNLP。以上海地铁10号线追尾事件为例进行网络舆情演化特征分析发现:①交通舆情演化迅速,周期较短爆发集中;②情感整体倾向负面;③官方机构及时发声、对事故调查公开透明并采取有效措施是缓解舆情较好手段;④大众对“卡斯柯”和“轻度”这2个词的关注贯穿事件始终。Transportation is related to domestic economy and livelihood of people.Perception of public opinions is necessary for government to manage traffic situations.In the era of Internet,access and analyze public opinions on traffic incidents from the internet are characterized by abundant data,objective information,and can be accessed in time.Based on electronic information sources such as Weibo,WeChat,and news client,features of public opinions on traffic incidents can be extracted.A systematic analysis method is developed to analyze online public opinions on traffic incidents.Major steps of it including obtain data and preprocess text,construct a classification model of traffic incidents,establish a calculation method of evaluation importance,analyze evolution trend of public opinion considering relevant life cycle,establish a database of sentiment on traffic incidents,analyze evolution of emotion states using machine learning,and coupling online information and visual analysis technology through the key words.The F-value of evaluation results of text classification is higher than 80%,and the accuracy of emotional polarity judgment is higher than that of widely-used SnowNLP.Taking a rear-end incident of Shanghai metro line 10 as a case study,the results show that:①public opinions of traffic incidents rapidly evolves within a short period,and intensively erupts;②emotions tend to be negative overall;③authority states official investigation and take effective measures in time is a better way to alleviate public opinions;④“Casco”and“Mild”are two keywords that public concerns about throughout this event.

关 键 词：交通工程 交通事件 舆情特征 主题分类 舆情演化 情感状态 

分 类 号：U116.1[交通运输工程]