基于朴素贝叶斯分类与群智感知实现对传染病与相应舆论的实时监控

Building Real-Time Monitoring of Infectious Diseases and Corresponding Public Opinion Based on Naive Bayesian Algorithm and Crowd Sensing

摘要: [背景/意义]将微博用户看作感知单元，形成群智网络，从公众情感角度对传染病的分布、传播与发展进行简洁直观、可视化的描述。[方法/过程]研究分五个部分。首先，爬取与事件相关的微博帖集；其次，利用数据过滤模型α过滤与事件关联度不高的帖集；同时，基于OpenStreetMap查询并检测位置以返回真实位置集；然后，基于Python的JIEBA分词系统，进行分词和词频处理，并基于朴素贝叶斯算法实现对关键词的情感预测，得到情感集；最后，通过情感反馈模型μ与疫情监控模型э产生情感曲线与疫情地图，并进行案例研究。[结果/结论]通过这种方法(内含四个模型)，可以从微博上生成传染病事件的疫情情感地图与公众情感曲线。其中，地图展示了各位置集的疫情严重情况与传播情况，情感曲线给出了公众反馈随时间的变化。研究发现，病例数越多与人口流动性较大的地区情绪表现更为消极，优良的医疗条件和积极的社交媒体宣传对公众情绪有一定积极效果。

Abstract: [Background/Meaning] The Weibo users are regarded as perceptual units to form a Crowd-sensing network, to produce intuitive and visual description of the distribution, transmission and devel-opment of infectious diseases and corresponding public emotions. [Method/Process] The study is divided into five parts: first, crawl the microblog posts associated with the event; second, use the data filtering model to filter the set of posts that have low correlation with the event; meanwhile, query and detect the position based on OpenStreetMap to return to the real position set; then, based on JIEBA word segmentation system of Python and Bayesian algorithm, achieve keyword emotional prediction; finally, through the emotion feedback model μ and epidemic monitoring model э, generate emotional curve and epidemic map, and make a case study. [Result/Conclusion] By this method (containing four models), epidemic emotion maps of infectious diseases and public emotion curves can be generated from Weibo. Among them, the map shows the epidemic situation and spread of each location set, and the emotion curve gives the public feedback changes. The study found that the more cases and the higher flow of population, the more negative the emotional performance. Excellent medical conditions and positive social media campaigns have a positive effect on public emotion.