Formulation of a Multi-relational Data-mining Method by Bottom-up Approach

本研究项目聚焦于多关系数据挖掘，涉及以下几个方面：(1)自下而上模式枚举的完备性；(2)过程的完备性和效率之间的权衡；(3)在化学性质鉴定、药物设计和语料库分析等各个领域的适用性。此外，本文还从理论和实践两个方面阐明了该方法与图挖掘等其他方法的区别。基于提出的基本属性和属性的组合运算，设计了一种高效的算法及其变体。该算法作为一个挖掘系统实现，并在各种类型的应用中进行了评估

This research project focused on multi-relational data-mining concerning (1) completeness of enumeration of patterns using the bottom-up approach, (2) trade-off between the completeness and efficiency of the procedure and (3) applicability in various areas, such as identification of chemical properties, drug design and analysis of language corpus. Furthermore, we also clarified differences of the approach from other methods including graph mining approach from theoretical and practical aspects.We designed an efficient algorithm and its variations based on basic properties that we proposed and combination operations for the properties. The algorithm was implemented as a mining system and evaluated in various types of applications