Based On The Matrix Of Weighted Association Rules Mining Algorithm

Since the 1960s, the database technology is used widely in many fields such as governments, business and scientific organizations etc. With the development of Internet technology, the data of these fields are rising up in the type of geometric explosion. Therefore, the requirement for information extraction of massive data is becoming increasingly urgent. The limitation of traditional database technology is that it can only query and search the database, but it cannot extract the knowledge from the database. As a result, plenty of knowledge cannot be explored and used effectively. Data Mining was created for solving this issue. Data mining is expected to find out the hidden rule or association among large amounts of accumulated data, and help the decision making and implementation in an effective way.The association rule mining which is one of the main research directions of data mining is used for finding out the relationship among data, and is a reference for decision-making. So far, the most famous Boolean association rule mining algorithms is Apriori algorithm which is proposed by R. Agrawal. However, there are two main problems for those association rule algorithms which are based on the Apriori algorithm. It scans databases frequently and ignores the items which have low frequency but with high value. Apriori algorithm thinks all the items have the same functions and importance, which is saying that it treats all items as the same priority under the algorithm's model. In order to operating the data mining at a more reasonable way, weight was brought as a new concept. It's required to study weighted association rules by using what we have learnt.In this thesis, an improved weighted association rule mining algorithms which is based on matrix is proposed. The comparison between the new algorithm and traditional weighted association rule mining algorithm is made by performance experiments. The experimental results show that the new algorithm has a better performance than traditional weighted association rule mining algorithm. The new feature of this improved algorithm is using matrix. The benefit is that it scans the database only once, at the meanwhile; it does not ignore the items with low frequency but high value, and what's more, it's using K support expectation as the reference of prune step, therefore, it solves the problem that is weighted frequent itemsets does not have anti-monotonicity in algorithms of weighted association rule comparing with in normal association rules.