Research Of Association Rules Algorithm Based On Multiple Tables For Relational Database

Database technology has become the technology which is important for mass data organizing and managing in information society with the rapid developments of database technology and widespread of Internet. The new barcode technology adopted in supermarket, bookstore specifically provides even more help on collecting data which is fast and convenient for enterprise. Data mining technology happens to reveal the knowledge hidden behind data from the mass data which can help people better understand the nature of object.Association rule is an important knowledge type in data mining. The algorithm studying on the association rule of relational database gives a wide development prospect. Currently, the regular algorithm is based on single table on this aspect, even though these algorithms can transfer to multiple tables, corresponding transformation is needed which is transforming multiple tables into one. In addition, other algorithms are proposed, such as the algorithm based on ILP which have some flaws in mining efficiency, field of application and mode which request new association rule mining algorithm being put forward to adopt requirement.After giving analysis and utilization of the advanced ides of traditional Apriori association rule algorithm, combined the Tuple ID Propagationt concept in CrossMine algorithm, this dissertation put forward a novel algorithm BMM applied on association rule mining among multiple tables in relational database based on the analysis of the time complexity of this algorithm and the restriction that can only be applied on transactional database. After the characteristics analysis of multi-value and multi-dimensionn, measures are taken to solve these according to these characteristics: data are preprocessed in relational data set to adjust the finding of association rule, then the object relation ID corresponding to each property is obtained through extending the Tuple ID Propagation concept, which is able to connect the properties among multiple inspected tables using object relation ID, which acts directly on multiple tables, eventually, chain list structure is used as data structure in the algorithm to indicate frequent itemset and related object relation ID to reduce the database accessing timesAt the end of this dissertation, the experiment is performed to verify BMM algorithm. BMM algorithm and the algorithm based on SQL are compared and analysis is given which shows that BMM algorithm owns shorter executing time.