道路交通伤害网络中文文本大数据关键挖掘方法及主要应用模式研究

网络文本大数据研究能提供常规道路交通伤害监测系统未采集但又有价值的补充信息，有助于更早发现新出现的道路交通伤害问题（如：快递员相关伤害）。本项目探讨当前道路交通伤害网络中文文本大数据挖掘研究面临的3个关键方法学挑战：①从包含大量噪音的数据中准确定义道路交通伤害相关信息；②从庞大的数据中自动和准确提取所定义的信息；③确定网络中文文本大数据在道路交通伤害防控的主要应用模式。本研究基于项目组前期多项伤害文本数据课题建立的检索词扩展、信息提取、质量控制和文本数据分析等策略，综合伤害流行病学、“互联网+”和大数据等理论，探讨道路交通伤害网络中文文本大数据的文本筛选、特征（变量）提取及分类标准，建立特征提取和文本分类的算法与统计模型（分类器），开发道路交通伤害大数据平台，系统地采集自2010年以来道路交通伤害防控相关的中文文本数据，探讨网络文本大数据的主要应用模式，为未来伤害大数据研究提供方法学参考。

Internet-based textual big data research can provide valuable data to supplement routine surveillance systems for road traffic injury and help detect emerging causes of injury earlier than possible with surveillance systems only (e.g. courier-related injury). This study aims to address three critical methodological challenges of internet-based textual big data mining in Chinese related to road traffic injury: (1) accurately extract road traffic injury related information from large and noisy datasets, (2) extract the defined and desired information about road traffic injuries from big data automatically and accurately, and (3) determine major application modes for internet-based textual big data in Chinese in the prevention and control of road traffic injury. Building from previous work for expansion of search words, extraction of information, quality control and textual data analysis by the research group, this study integrates underlying theories of injury epidemiology, “Internet +” and big data to develop accurate criteria for document screening in Chinese, extraction and classification of text-based features (variables), establishment of reliable algorithms and statistical models for feature extraction and text classification (classifiers), and building of a platform to systematically collect textual big data for road traffic injury in Chinese from 2010. Findings will readily extend to explore other application modes for big-data and offer methodological references for injury-related big data research in the future.