面向社交大数据的热点事件预测

项目名称： 面向社交大数据的热点事件预测

项目编号： No.61502246

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 其他

项目作者： 吴振宇

作者单位： 南京邮电大学

项目金额： 20万元

中文摘要： 社交网络是现实世界的映射。在社交大数据中预测热点事件，对发现用户群体、研究群体的活动规律，进而，抑制网络谣言传播从而维护信息安全，监控突发事件从而做出智能决策等方面具有重要意义。已有的研究主要分析文本内容，提取热点事件相关的关键词，统计关键词的频率预测热点事件。然而，社交大数据具有实时性、多样性及高动态性等特征，只处理内容预测热点事件相对比较困难。并且，已有的方法无法判断关注热点事件的用户群体，也就无法预测小众范围内的热点事件，个性化不明显。实际上，用户产生的内容通过社会交互相互关联，并且群体行为具有持续性。本课题融入人的认知以及由社会交互形成的群体智能，使用认知物理学的方法处理非结构化数据，研究热点事件的发展与群体智能强相关，与具体事件的语义弱相关的规律，以达到热点事件预测的目的。相关研究成果可以为人工智能的其它领域，比如机器人、智能驾驶的研究提供思路。

中文关键词： 社交大数据；认知计算；群体智能；拓扑势；事件检测

英文摘要： Social network is the reflection of the real-world. The prediction of popular events in social big data is meaningful for discovering the crowd and studying users' activities. Furthermore, the cyber security can be guaranteed by inhibiting the spread of rumors, and intelligent decisions can be made by monitoring emergencies. Existing studies predict popular events by analyzing user generated contents to retrieve the key words about the events. However, the social big data has characteristics such as real-time, variety, and velocity. Thus, it is difficult to predict popular events only by means of content analysis. Moreover, existing methods cannot identify the crowd who focuses on the events. This shows that existing methods lack the personalization. In fact, user generated contents are connected by social interactions, and the crowd behavior has sustainability. This project incorporates the cognitions of human and the collective intelligence formed by social interactions, and uses the cognitive physics methods to process the unstructured data. Furthermore, the popular events are predicted by studying how the evolutions of popular events are strongly related with the collective intelligence, while are weakly related with the semantics of the events. The research achievements can be used to other domains of artificial intelligence, such as the robot and the intelligent driving.

英文关键词： Social big data;Cognitive computing;Collective intelligence;Topological potential;Event detection