网络舆情监控算法研究与分析  

Research and Analysis of Network Public Opinion Monitoring Algorithms

作　　者：谢卫红[1,2] 杨超波 朱郁筱 李忠顺 蒋瞰阳[3] Xie Weihong;Yang Chaobo;Zhu Yuxiao;Li Zhongshun;Jiang Kanyang(School of Management,Guangdong University of Technology;Institute of Big Data Strategic Research,Guangdong University of Technology,Guangzhou 510520,China;College of Computer Science and Engineering,Rochester Institute of Technology,New York 10041NY212,USA)

机构地区：[1]广东工业大学管理学院 [2]广东工业大学大数据战略研究院,广东广州510520 [3]罗切斯特理工大学计算机科学与工程学院,美国纽约10041NY212

出　　处：《科技管理研究》2019年第22期197-205,共9页Science and Technology Management Research

基　　金：国家自然科学基金项目“大数据背景下的网络隐私顾虑影响因素及行为效应研究:基于多维发展理论视角”(71672043)

摘　　要：梳理现有的网络舆情监控算法,归纳为经典算法和拓展算法,分析算法的优缺点并对比常见算法的性能,整理部分专家学者对这些监控算法的优化改进研究情况并分析其研究成果。从技术、管理和应用等3个角度评价现有网络舆情监控算法的优化改进环节和取得的成效,并指出其存在的问题主要集中在只基于文本、针对单一数据类型、没有考虑事件和用户的差异化特点和动态变化情况、缺乏综合监控体系思维和管理机制等方面。进而从网络舆情的特点、发展规律、驱动因素、现有监控算法不足和监控效果期望等角度探讨未来网络舆情监控算法的发展趋势。This paper combs the current network public opinion monitoring algorithms, summarizes them into classical algorithms and extended algorithms, analyzes the advantages and disadvantages of these algorithms, also compares the capabilities of famous algorithms, arranges the research situation of some experts and scholars on the optimization and improvement of these monitoring algorithms and analyzes their research results. It evaluates the optimization and improvements of these monitoring algorithms from the perspectives of technology, management and application. It is pointed out that the problems are mainly based on text, aiming at a single data type, without considering the differential characteristics and dynamic changes of events and users, and lacking comprehensive monitoring system thinking and management mechanism. Next, the development trend of network public opinion monitoring algorithm in the future is discussed from the point of view of the characteristics, development law and driving factors of network public opinion, the deficiencies of existing monitoring algorithms, and the expectation of monitoring effect.

关 键 词：网络舆情 监控算法 文本分类 文本聚类 情感倾向 

分 类 号：C93[经济管理—管理学]