聚类集成因其融合多聚类结果以得更优聚类的能力，近年来已成为聚类分析领域的一个热点研究方向。但是，当前聚类集成算法往往针对普通规模的数据集而设计，且大多仅考虑单源场景，难以应对现实数据经常存在的多源性与大规模性特点。在多源大规模数据场景下，目前聚类集成研究仍面临多源建模、高阶融合、高效运算等若干亟待解决的关键问题。针对于此，本项目拟以聚类集成与二部图模型为切入点，研究基于异构二部图模型的多源大规模数据聚类集成新框架，着重开展三个方面的理论研究，分别是：1）多源大规模二部图高效构建及其高阶融合；2）基于两类异构二部图的无监督多源大规模聚类集成；3）基于非平衡类标传播的半监督多源大规模聚类集成。进一步，本项目拟开展所建立算法在多源监控视频数据、多源癌症基因数据以及多源社交网络数据上的应用研究。本项目工作将丰富数据挖掘与大数据分析的理论与方法，特别是推动大规模聚类与聚类集成研究的深入发展。

Due to its ability of combining multiple clusterings into a better clustering result, the clustering ensemble technique has in recent years become a popular research topic in the area of clustering analysis. However, most, if not all, of the existing clustering ensemble algorithms are designed for moderate-scale and probably single-source datasets, which unfortunately lack the desirable ability to handle multi-source large-scale datasets in real-world scenarios. When dealing with multi-source large-scale datasets, the current clustering ensemble research is still faced with several crucial problems that remain to be solved, such as the multi-source modeling problem, the higher-order integration problem, and the computational efficiency problem. Inspired by the clustering ensemble and the bipartite graph model, in this project, we plan to conduct research on clustering ensemble algorithms for multi-source large-scale data based on heterogeneous bipartite graph models. In particular, our theoretical research mainly focuses on three sub-topics, namely, (1) the efficient construction and higher-order integration of multi-source large-scale bipartite graphs, (2) unsupervised clustering ensemble for multi-source large-scale data based on two types of heterogeneous bipartite graphs, and (3) semi-supervised clustering ensemble for multi-source large-scale data based on imbalanced label propagation. In terms of application research, we plan to apply our newly designed clustering ensemble algorithms to analyze the multi-source visual surveillance datasets, the multi-source cancer gene expression datasets, and the multi-source social network datasets. The research work of this project will enrich the theory and methodology in the fields of data mining and big data analytics, and particularly promote the development of the large-scale clustering and the clustering ensemble research.