Characteristic relational patterns

摘要: Research in relational data mining has two major directions: finding global models of a database and the discovery local patterns within database. While show how attribute values co-occur detail, their huge numbers hamper usage analysis. Global models, on other hand, only provide summary different tables attributes relate to each other, lacking detail what is going at level.In this paper we introduce new approach that combines positive properties both it provides detailed description complete using small set patterns. More particular, utilise rich pattern language can be encoded by such Then, based MDLprinciple, novel RDB-KRIMP algorithm selects allows for most succinct encoding This set, code table, compact terms We resulting very small, size number its patterns: reduction up 4 orders magnitude attained.