The Research On The Fast Class Association Rules Algorithm

With the development and extensive application of modern technology of database, the quantity of data and the complexity are increasing sharply in database. It is necessary to describe and discover the information and the relationships concealed in these data with a new technology.Data mining, which is a method of extracting previously unknown and potentially useful patterns and relationships from data,is widely used in business management,production control,market analysis, projects design and science exploring,etc.The relationships between patterns in database are explained by the rule that is an importmant field in data mining.Generally,the rule is divided into two categories. One is association rules which connect with problems of market basket in transaction database, and the other one is classification rules which relate with prediction in relational database. In this dissertation,we will mainly discuss class association rules that mine classification rules with association rules algorithms.Relational databases are often denser than transaction databases,so mining on them for class association rules may be difficult.For example: surplus redundant rules.Therefor, this paper introduce an optimal class association rule mining algorithm, is named OCARA. Since OCARA uses optimal association rule mining algorithm, and the rule set is sorted by priority of rules, it results in a more accurate classifier. This algorithm is compared with C4.5, CBA, RMR on UCI eight data sets respectively. Experiment results show that it has better performance.In addition,the scale and time of mining class association rules will increase sharply in incomplete relational database when it is higher dimension.In this paper, a rough set theory based fast ORD algorithm for mining association rules is presented. It uses the attribute reduction algorithm to reduct attributes firstly. Then the fast and efficient algorithm for pruning redundant itemsets and redundant rules, ORD algorithm is applied to obtaining association rules. Experiment results show it has good performance.