Minimum description length principle: generators are preferable to closed patterns

摘要: The generators and the unique closed pattern of an equivalence class itemsets share a common set transactions. are minimal ones among equivalent itemsets, while is maximum one. As generator usually smaller than in cardinality, by Minimum Description Length Principle, preferable to inductive inference classification. To efficiently discover frequent from large dataset, we develop depth-first algorithm called Gr-growth. idea novel contrast traditional breadth-first bottom-up generator-mining algorithms. Our extensive performance study shows that Gr-growth significantly faster (an order or even two orders magnitudes when support thresholds low) existing mining It can be also state-of-the-art itemset algorithms such as FPclose CLOSET+.