伴随着互联网、个人传感设备（手机等）、微信、微博等信息技术的飞速发展，人类社会产生了数以亿计的具有非结构化、多态异构、多源异质、不完备以及过度冗余等特性的混杂数据，如何从这些海量的数据中挖掘出有用信息是当前信息领域亟待解决的核心问题之一。. 围绕海量低质、多态异构、多源异质的混杂数据，本课题拟开展混杂数据模式识别的基础理论与方法研究，包括海量低质混杂数据的清洗与约减、多态异构混杂数据的同构表征、多源异质混杂数据的有效性辨识以及多源混杂数据的关联建模等，实现混杂数据的清洗、统一表示、有效性辨识、进而实现数据间的关联建模，构建混杂数据模式识别的基础理论，并将理论研究成果应用于具有混杂特性的敏感内容（暴恐视频等）的分析与挖掘中。. 研究成果将丰富模式识别的基础理论，并为社交媒体分析，多媒体应用，大数据及敏感内容的检测等应用提供有力的理论和技术支撑。

As the rapid development of Internet、personal sensor equipment (such as mobile phone)、wechat and microblog, vast amount of diverse data which are unstructured, low quality, multi-modal, heterogeneous and non-homologious，have been generated. How to mine some useful information from the massive data is a key technology nowadays.. Focused on the diverse data which are low quality、multi-modality and heterogeneous、multi-source and non-homologious，the project will study the pattern recognition theory and method for the massive diverse data, including the cleaning and reduction of low-quality and redundant diverse data, consistency representation of heterogeneous data, the validity identification of multi-source and non-homologous diverse data, association modelling of diverse data, etc. The project will further apply the theory and method to the analysis and mining of sensitive content with diverse data properties.. The research achievement will not only enhance the theory of traditional pattern recognition, but also provide theoretical and technological support to the applications such as social media analysis, multimedia applications, big data anakysis and sensitive content detection.