面向电信用户群体交互大数据的事件线索在线涌现与预判研究

With the quickly development of the mobile internet and smart terminal, telecom network becomes important carriers for social activities and information diffusion, and records various communities of more hundreds of millions peoples in real-world. Those come about more and more telephone network crime. It is important research topic that how takes opportunities and challenges in social crisis management. The opportunities and challenges must be develop new techniques of online emerging and pre-estimating even clue from big data of telecom users group social activities. .Our task will establish evolution representation theory for large-scale user group interaction, and breakthrough on key science problems such as interaction knowledge representation, online emerging and pre-estimating of importance even clue. Base on key science problems, we firstly define an evolution representation theory of telecom user group dynamical interaction model that lays the theoretic foundation for even clue online emerging and crisis pre-estimating. Secondly, we will study key techniques of multi granularity even clue online emerging and crisis pre-estimating. Lastly, we develop the prototype system to verify our key technology and theory..Our research results are expected to be efficient not only in solving problems that online generates multi granularity even clue model from large-scale telecom user group interaction information. Its potential in improving the accuracy and efficiency of crisis pre-estimating is theoretically significant. Meanwhile it has a promising prospect in defusing the tensions over national telephone network crime issues as well as ensuring the rapid development in national economy.

随着电信移动互联网及智能终端快速普及，电信网络成为人类社会关系维系及信息传播的重要载体、真实记录数亿用户的社会关系和言行信息，引发电话网络犯罪愈演愈烈，如何应对严重挑战成为社会风险防控领域关注焦点，迫切需要开展面向电信用户群体交互大数据的事件线索在线涌现与预判研究。本项目旨在建立用户群体动态交互的演化表示理论，突破国计民生的事件线索涌现和风险预判技术瓶颈，解决用户动态交互演化表征、事件线索在线涌现等问题。研究内容：首先，研究电信用户群体动态交互的演化表示模型，奠定事件线索涌现和风险预判的理论基础；其次，研究用户群体交互中事件线索在线涌现和风险预判算法；最后，开展实例验证，验证提出的理论和关键技术。通过本项目研究，尝试从科学角度回答能否从海量用户交互信息中在线生成影响国计民生的事件线索，具有重要的理论意义；对于缓解我国在电话网络犯罪领域面临的严峻形势和保障国民经济快速发展具有广泛的应用前景。

本项目针对引发电话网络犯罪愈演愈烈，如何应对严重挑战成为社会风险防控领域研究热点，开展了面向电信用户群体交互大数据的事件线索在线涌现与预判研究。主要研究内容包括：1）研究电信用户群体动态交互的演化表示模型，奠定事件线索涌现和风险预判的理论基础；2）研究突破用户群体交互图谱在线构建、事件线索在线涌现和风险预判等三大关键技术瓶颈；3）选取电信网络空间的群体交互数据，开展应用验证。.本项目经过4年研究，取得了4项重大研究成果：首先提出了群体动态交互的三层演化网络表示理论方法，开创了网络空间多层次、多尺度属性和介尺度复杂性表征理论方法，构造出了人类社会与信息空间动态映射超大网络表示模型；其次突破了大规模群体交互图谱在线构建、事件线索在线涌和事件线索风险预判等3项关键技术瓶颈，形成了一套完整电信网络空间的大数据标准化治理方法体系。最后选取了2017年全国电信移动用户的全年信令协商短文本结构关系型数据（约500TB）作为训练样本空间，形成了1套网路反诈骗主动巡检与预警系统。从中主动甄别出了150万条可疑网络电话诈骗线索，经过专家和历史信息主动比对，甄别出了145万条真实存在网络电话诈骗记录，准确率达到91.5%左右。为我国快速有效遏制网络电话全境蔓延起到决定性作用，帮助国家挽救了广大人民财产损失。