Associative Classification Algorithm And Its Application In Electronic Business Recommendation System

Classification and association rules discovery are important research fields in data mining and they have gained perfect research achievement. As a novel data mining subject, associative classification utilizes association rules discovery to build classification system. Associative classification algorithm uses rules discovery method to extract classification rules from classification data sets, prunes rules by means of pruning mechanism and builds classifier based on some rules.Because it mines strong association between data attributes and class label, associative classifier could achieve high accuracy. However, in rules generation period, associative classification could generate huge amount of rules and algorithm would need large overhead to process them. Consequently, it is necessary to generate a rules set with a small scale and have an effective pruning method. In this paper, a associative classification algorithm based on ECLAT is proposed and it combines the property of compact rules set and the feature of ECLAT. The proposed method introduces pruning strategy during rules mining period, tests the confidence of generated rules and deletes redundant rules and equal class. The experiment shows that this method could dramatically compresses the rules number of original set, shortens the running time and reduces the overhead.In association rule mining and associative classification, interestingness measure is introduced to extract rules and reduce the number of rules. In this paper, some common interestingness are introduced and an experiment compares the classification impact of different interestingness on different data sets. The experiment shows that interestingness could reduce the number of rules while no interestingness could achieve optimal result for all data sets.Lastly, associative classification is used in electronical recommendation system. In P2P site Beiyangyuan PT, recommendation algorithm uses associative classification to learn classification modle from users’records, builds associative classifier, and classifies new users. Algorithm recommends matched resources to users based on their class and this application shows the practical significance of data mining.