基于模型集成的突发事件舆情分析与趋势预测研究

Public sentiment of emergent events based on model integration and its trend prediction

近年来,突发事件发生后,事件演进受网络舆情的影响越来越大,分析突发事件发生后网民情绪并进行预测,可为政府部门的应急管理和策略设计提供有效支撑,赢得宝贵的时间.本文提出了一个基于模型集成的微博情感分析与预测模型,对突发事件微博舆情进行情感分类与趋势预测.为了更准确地分析微博情感与未来走势,首先,利用多模型集成策略对突发事件相关的单条微博进行情感分析;接着,将单条微博情感进行集成,形成微博情感时间序列;再次,利用多模型集成思路对微博情感的未来走势进行预测;最后,通过实例验证提出方法的有效性.实证结果表明,集成模型较传统分类在微博情感分析上具有优势,集成模型较传统回归模型在微博情感走势预测同样具有明显优势,可以取得较高的预测精度.

In recent years, after the occurrence of emergent events, the event evolution is affected by network public sentiment seriously. Analysis and prediction of network public sentiment can provide effective support for the government＇s emergency management and strategy design, which can win precious time for the government. In this paper, a model is proposed for microblog sentiment analysis and forecast model based on model integration, which can offer emotion classification and trend prediction for microblog public sentiment of emergent events. In order to analyse microblog sentiment and its future trend accurately, firstly, multi model integrated strategy was used to analyse a single microblog sentiment related to emer- gent events. Secondly, these single microblog sentiment were integrated to form microblog emotional time series. Thirdly, the future trend of the microblog emotion was predicted with integrated modeling method. Finally, validity of method was proposed through the example verification. The empirical results show that comparing with traditional classification method, the integrated model has its advantage in microblog emotion analysis. And the obvious advantages of the integrated model also showed in trend prediction for microblog public sentiment of emergent events, which showed a high prediction accuracy.