In recent years, knowledge discovery in database receives the artificial intelligence and widespread value in database field. Mining association rules is an important topic of data mining research. Its main research object is the transaction database, its essential target is to discover whether there is certain connection relation between items in transaction database.The present judgment criteria of association rules are a support and a confidence .If the association rules are generated according to the criteria, a lot of them are invalid and false.Presently the formalize definition of association rules is adopted the formation of X(?)Y, there into X as to precondition,Y as to subsequence. But actually we only know whether Items be sold together.We don't know whether X induce Y or Y induce X.So the formation of X(?) Y is not fit to the fact,it may lead to wrong results.We carry on carefully analysis to classic algorithms like Apriori, FP_gworwth algorithm and so on, and offered their improvement forms. To reduce invalid rules in mining association rules, we have analyzed the reasons and presented to add the effect or the relative confidence, we classify strong association rules into positive, invalid and negative association rules. We offer improving algorithm of mining association rules with new judgment criterion.We studies the existing judgment criteria of data mining, and propose a new formation of the definition of association rules. And we offered new algorithm.The results of test shows that the method we proposed can redude invalid rules obviously. |