An effective approach to mine relational patterns and its extensive analysis on multi-relational databases

Abstract

The real world applications of data mining necessitate more complicated solutions when the data includes large quantity of records in several tables of relational database. One of the possible solutions is multi-relational pattern mining, which is a form of data mining applicable to data in multiple tables. In this paper, we have developed an effective approach to mine relational patterns from multi-relational database. Initially, the multi-relational database is represented using a tree-based data structure without changing their relations. A tree pattern mining algorithm is devised and applied on the constructed tree-based data structure for extracting the frequent relational patterns. Experimentation is carried out on two different databases and the results are compared with the previous approach using number of similar relational patterns generated and the computation time. The comparative analysis shows that the proposed approach is more effective in mining performance and computation time compared to the previous approach.