移动与社交联线大数据的高效协同分析理论及应用

Due to the proliferation of mobile internet and the trend of miniaturization, multimedia, networking and multi-functionalization of mobile devices, billions of mobile devices produce a huge amount of behavior sensing data and online social media data in every second. These data truly describe human behaviors and society via the web platforms from online to offline, from subjective to objective, from ego to community, from society to nature, from behavior to event, and from object to scenery. It has also brought great challenges to research on knowledge discovery and management technologies. Traditional analytics models on single type, single task and small sized data cannot well adapt to the urgent needs from online-offline big data analysis. To address this problem, this project proposal aims to discover relations between offline mobility sensing information and online social media. Specifically, we study methodologies to extract attribute information from heterogeneous data sources, and construct fusion models for multi-source information. We present high performance information fusion and collaboration framework in order to discover the semantics and knowledge from multi-source heterogeneous big data, and thus we are able to conduct collaborative prediction, correlation analytics and causal analytics. We also study techniques for user preference database and personalized recommendation to satisfy the urgent needs of content recommendation/marketing and routing services from the big data application perspective.

随着移动互联网的普及，移动设备小型化、多媒体化，网络化和多功能化的趋势，数以亿计的各类移动设备每时每刻都在产生大量的行为感知数据和大量的网络媒体数据，从线上到线下，从主观到客观，从自身到群体，从社会到自然，从行为到事件，从物体到场景，通过互联网平台的收集和呈现，真实反映了人类社会生产生活的形态万千的信息，也给知识挖掘和管理技术研究带来了重大的挑战。传统的针对单一类型、单一任务和面向小数据量的分析方法已经不能满足迫切地现实需要。为了克服这个挑战性难题，本研究旨在建立线下移动感知信息与线上社交媒体之间的有效联系，从多个平台中协同抽取属性信息；构建多源信息融合方法，设计移动和社交的高效信息融合和语义协同框架；挖掘多源异构联线大数据当中的语义和知识，进行融合预测、关联分析和因果推断等，建立海量用户的属性和偏好数据库，实现深度有效的个性化需求分析，构建精准的内容营销及出行规划等推荐系统和大数据应用。

该项目按期按计划完成，建立了线下移动感知信息与线上社交媒体之间的有效联系，从多个平台中协同抽取属性信息；构建了多源信息融合方法，设计移动和社交的高效信息融合和语义协同框架；挖掘多源异构联线大数据当中的语义和知识，进行融合预测、关联分析和因果推断等，建立海量用户的属性和偏好数据库，实现深度有效的个性化需求分析，构建精准的内容营销及出行规划等推荐系统和大数据应用。共发表高质量论文20篇，申请或已获批发明专利4项，培养研究生11名（博士3位，硕士8位）。