Multifaceted exploration of nonhomogeneous and ambiguous data by combining partial similarities

由于工业和科学数据的计算机化，我们可以自动收集或系统地获取巨大的数据集。虽然这些数据比以前更容易获取，但当我们试图对它们进行统计分析和利用时，它们的劣质会带来问题。由于使用目的不明确且事先不固定，许多级别的信息被折叠成一个单一的数据集，因此，它们的质量不够同质。为了解决这个问题，我们基于最近出现的子结构枚举技术开发了新的统计方法，它可以通过结合数据中的部分或局部相似性来分析这些类型的数据。

Due to the computerization of industrial and scientific data, we can automatically collect or systematically obtain a huge dataset. Whereas these data are more easily accessible than before, their poor quality causes problems when we try to statistically analyze and utilize them. Many levels of information are collapsed into a single dataset since the purpose of use is ambiguous and unfixed in advance, and as a result, their quality is not sufficiently homogeneous. To address this problem, we developed novel statistical methods based on recently-emerged techniques for substructure enumeration, which can analyze those types of data by combining partial or local similarities in the data.