在生物信息学的基因表达分析领域，通过挖掘保序子矩阵从数据中发现了有用的生物关联，成功实现了从数据到知识的转化。然而，由于保序子矩阵挖掘问题本身的计算复杂性，对于该问题的研究大多局限于针对中小规模数据的分析。在本课题中，我们考虑在更广泛应用领域针对不同特征数据的分析需求，比如，从大规模的稀疏数据中发掘关联，以及发掘数据中存在时间延迟的关联等。首先，我们通过扩展保序子矩阵模型，试图刻画不同应用领域数据中存在的有用保序关联。其次，针对每一种扩展模型，我们研究与设计相应的挖掘算法以实现子矩阵的有效挖掘。最后，针对日益重要的大数据应用分析需求，我们设计与构建保序子矩阵挖掘的分布式计算平台以实现在超大规模数据上的子矩阵挖掘。我们基于本课题的研究工作在保序子矩阵挖掘领域将形成一套完整的体系，可以应对不同领域、不同特征数据的分析需求。

In the field of gene expression analysis in bioinformatics, through mining order-preserving submatrices (OPSM) from the gene expression data, biologically significant correlations can be discovered, which successfully achieves the conversion from data to knowledge. However, due to the inherent computational complexity of the OPSM mining problem, most of the research work on this problem is restricted to the analysis of small- to middle-scale matrix data. In this project, we consider challenging data analysis tasks caused by different data characteristics in wider application areas, such as the need for finding correlations from large-scale sparse data, the need for finding correlations with time lagging, and etc. We first try to improve the original OPSM model to well capture the useful order-preserving correlations in these application data. Then, for each improved OPSM model, we design and develop efficient pattern mining methods. Last but not the least, in order to carry out efficient data analysis in Big Data applications, we construct a distributed computing platform for mining OPSM patterns from extremely large-scale data. Our research work in this project form a unified and complete system for OPSM mining, which can meet the needs for data analysis from diverse application areas.