Mining closed and maximal frequent embedded subtrees using length-decreasing support constraint

This paper presents algorithm SCCMETreeMiner which can find all closed and maximal frequent embedded subtrees using length-decreasing support constraint. SCCMETreeMiner combines the rightmost path expansion scheme and projection technique to construct pattern growth space, and uses several techniques proposed in the paper to prune the branches of the enumeration trees that do not correspond to closed or maximal frequent subtrees under length-decreasing support constraint. The performance of the algorithm is studied through extensive experiments by using various length-decreasing support constraints and datasets. All experimental results show that our algorithm is effective and efficient, and it generates more concise result set, which is irredundant and interesting to users.