Knowledge Discovery from Structured Data with Efficient Methods of Searching Lower Sets Based on Frequency

本研究项目旨在从存储在数据库中的大规模和各种数据类型的数据中构建有效的知识发现基础。我们阐明了由各种类型的结构化数据组成的晶格结构的各种特性，通过在晶格中有效地搜索下集来设计机器学习/知识发现算法，并将其中一些算法应用于理论开放问题或信息学以外的科学问题。我们发现下集作为闭集的性质在基本理论和应用上都是有用的。

This research project aimed at constructing foundations of efficient knowledge discovery from data of large scale and of various data types, stored in databases. We clarified various properties of the lattice structure consisting of various types of structured data, designed machine learning/knowledge discovery algorithms by using efficient search of lower sets in the lattices, and applied some of them to theoretical open problems or to problems in science other than informatics. We found that the property of lower sets as being closed sets is useful for both fundamental theories and applications.