Mining Frequent Rooted Ordered Tree Generators Efficiently

Abstract

With the wide applications of tree structured data, such as XML databases, research of mining frequent sub tree patterns have recently attracted much attention in the data mining and database communities. Due to the downward closure property, mining complete frequent sub tree patterns can lead to an exponential number of results. Although the existing studies have proposed several alleviative solutions (i.e. mining frequent closed sub tree patterns or maximal sub tree patterns) to compress the size of large results, the existing solutions are not suitable some real applications, such as frequent pattern-based classification. Furthermore, according to the Minimum Description Length (MDL) Principle, frequent rooted sub trees generators are preferable to frequent closed/maximal sub tree patterns in the applications of frequent pattern-based classification. In this paper, we study a novel problem of mining frequent rooted ordered tree generators. To speed up the efficiency of mining process, we propose a depth-first-search-based framework. Moreover, two effective pruning strategies are integrated into the framework to reduce the search space and avoid redundant computation. Finally, we verify the effectiveness and efficiency of our proposed approaches through extensive experiments.