Higher-Order Logic-Based Knowledge Representation and Clustering Algorithm

With machine learning and knowledge discovery deeply development and breadth of the expansion in many areas of application with the complex structure data, the knowledge discovery of the complex structure data has become a core issue in the field of knowledge discovery and, the search space of patterns during the course of knowledge discovery is very large. Although the Inductive logic programming-ILP can flexibly express the multi-relation data,the background knowledge and the complex mode involving multi-relationship in process of relational learning and multi-relationship data mining, but the problems of predicate invention and utility are also difficult to solve and remain open problems in knowledge discovery based on first-order logic. The high-order logic knowledge representation can solve the problem effectively. For the higher-order logic knowledge representation formalism-Escher can express all kinds of complex structured data. It not only can provide strong guidance on the search for frequent patterns with its strong typed syntax but also can resolve the problem of the invention of new predicates with its higher-order characteristic. It is fit for knowledge discovery in complex structured data. This paper will investigate the knowledge discovery in complex structured data by employing Escher as knowledge representation formalism. In the case of algorithms, clustering of complex structured data is studied in it and experimental verification of its effectiveness.