基于社交媒体大数据的交通感知分析系统  

A Traffic Sensing and Analyzing System Using Social Media Data

作　　者：郑治豪 吴文兵 陈鑫[3] 胡荣鑫 柳鑫 王璞[1] ZHENG Zhi-Hao;WU Wen-Bing;CHEN Xin;HU Rong-Xin;LIU Xin;WANG Pu(School of Traffic and Transportation Engineering, Central South University, Changsha 410075;School of Software, Central South University, Changsha 410075;School of Information Science and Engineering, Central South University, Changsha 410075)

机构地区：[1]中南大学交通运输工程学院,长沙410075 [2]中南大学软件工程学院,长沙410075 [3]中南大学信息科学与工程学院,长沙410075

出　　处：《自动化学报》2018年第4期656-666,共11页Acta Automatica Sinica

基　　金：国家自然科学基金面上项目(61473320);湖南省科技计划项目(2015RS4011);中南大学创新驱动计划项目(2016CSX014)资助~~

摘　　要：社交媒体数据中蕴含了丰富的交通状态信息,这些信息以人类语言为载体,包含了大量对交通状态的因果分析与多角度描述,可以为传统交通信息采集手段提供有力补充,近年来已成为交通状态感知的重要信息来源.本文以新浪微博为主要数据来源,分别利用支持向量机算法、条件随机场算法以及事件提取模型完成微博的分类、命名实体识别与交通事件提取,开发了基于社交媒体大数据的交通感知分析与可视化系统,可以为交通管理部门及时提供交通舆情及突发交通事件的态势、影响范围、起因等信息.在交通信息采集系统建设较为薄弱的地区,本文建立的系统可以为交通管理提供信息补充.Social media data, which encapsulate abundant traffic status information, have gradually become an important data source for sensing traffic status. The information recorded by human language contains a large amount of causality analysis and multi-angle descriptions of the traffic condition, acting as a powerful supplement to traditional traffic information collecting methods. Employing Sina Weibo as a main data source, we apply SVM algorithm, CRF algorithm and event extracting model for classification, named entity recognition and events extraction of microblogs. We develop a traffic sensing and visualizing system, which can collect public opinion, situations, scales and even origins of traffic incidents for transportation agency. Furthermore, this system can provide traffic information for the transportation department in the area which lack traffic detectors.

关 键 词：社会交通 机器学习 文本分类 命名实体识别 数据可视化 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论] TP391.1[自动化与计算机技术—计算机科学与技术]