Method Of Tendency Mining In Dynamic Association Rules Based On Associative Classification

Data mining is a research focus in KDD. Data mining is generally a process of searchingrules which have specific relation or verifying the known patterns from database or datawarehouse. Data mining model can be roughly divided into regression, association rule, timeseries model, cluster model and classification model in which association rule mining is animportant topic in data mining which is widely researched. At present the traditionalassociation rule mining is a kind of static mining for transactional database as is said that theydon’t think the rules will change over time. Considering that much rules mined in some realdatabase has time characteristics, it is necessary to add time factor in observing the changes ofassociation rules in time which process is called the dynamic association rule. During theprocess of mining dynamic association rules the time factor is interval divided to judging therules under the evaluation system of the support and confidence. The tendency mining indynamic association rules introduces the trend threshold to pruning useless rules to avoidgenerating invalid dynamic association rules mining. The association classification is amethod by introducing a set of training data with category identifier associated classifier toforecast the unknown data object. It has a higher classification accuracy and strongeradaptability. But the traditional association classification methods have many problems in thealgorithm performance, quality and efficiency as well as the classification of the pruningunderstanding.In this paper, the method of tendency mining in dynamic association rule based onassociation classification when combining the tendency mining and association rule togetheron the basis of relevant theoretical researches. The priority rule interestingness model and thecompatibility feature vector SVM model are proposed. The models are used to pruning theclass association rule sets generated by the method of tendency mining in dynamic associationrule to generate the final rules. The algorithm takes advantage of association rule and it has abetter effect in rule mining and predicting. By comparing the experimental data, it is provedto be a better applicability and higher accuracy.By analyzing the tendency of the rules, the method solves the matter of how to select thesupport vector in the process of dynamic association rule mining. Comparing with thedynamic association rule mining methods proposed in recent years, the method proposed inthis paper mined less rules and higher precision and effectively avoid the blindness ofdynamic association rule mining.