Research Of Method On Mining Valid Association Rules

Data Mining has become a research hotspot in many fields such as ArtificialIntelligence and Pattern Recognition in recent years; with the rapid increasing of dataquantity, technologies of data mining have received more and more attention. Miningassociation rules is one of the most active directions in data mining. In this paper,technologies of mining association rules are given a deep analysis and research. And themain work includes the following several aspects: 1 Now the research on association rules is focused on how to enhance its efficiencyrather than analyze the rules. In this paper, two problems about the criterion of evaluatingassociation rules and the relativity between the antecedent and the consequent are givenfirst, and then the related research about mining valid association rules is summarized. Anew frame, Support – match, for mining valid association rules is proposed in order toremedy the limitation of the traditional one which is enable to describe therelativity-questions of antecedent and consequent of the rule. Then an example is given tocompare the rules created by match frame with the ones created by support-confidenceframe. Results have shown that the new approach not only contains the high relativitybetween the antecedent and the consequent but also reduced the number of redundant rules.And some expanded applications of this new frame have been given with some results. 2 In real-world applications, the association rules are changing because the databasesor the data warehouses are changing along with time. Some research workers have takenup in how to update the association rules with high efficiency and some algorithms havebeen proposed. Updating the association rules mainly involves three aspects: Firstly, howto create db+DB assiociation rules when a new database db is added into database DB.Secondly, how to create the DB-db association rules when database db is deleted fromdatabase DB. And thirdly, how to update increments with weights when given the noveltyof the new database db. 2.1 First of all, this paper carries on the analysis and research to the first aspect andadvances a new approach of VFUP(Vector-Based Fast Updating Algorithm) and comparesthe new approach with the former one, which shows the high efficiency and the feasibilityof the new one. 2.2 Then, the problem of how to update the association rules when adding a newdatabase db without changing the minimal support and the minimal confidence is givenand analyzed. Anew algorithm of incremental updating based on weighing the itemsets hasbeen given combining the efficiency of VFUP and a new concept of match, considering thenovelty of the added database and the interesting of the rules. The new algorithm mines theDB+db association rules by weighing the frequency itemsets and the non-frequencyitemsets simultaneously in the added database db, which can enhance the effect that thefrequency itemsets and the non-frequency itemsets in database db pressed on DB+db. Inthe process of designing the algorithm of n- incremental updating, this paper intends todesign a high efficiency algorithm to increase the efficiency combining the property ofincremental updating based on weighing and analyzes the rules created by the new frame.Finally, this paper compares the algorithm of incremental updating based on weighing withthe one of incremental updating; the results have shown the efficiency of the new one.