Answering the Most Correlated N Association Rules Efficiently

摘要: Many algorithms have been proposed for computing association rules using the support-confidence framework. One drawback of this framework is its weakness in expressing notion correlation. We propose an efficient algorithm mining that uses statistical metrics to determine The simple application conventional techniques developed not possible, since functions correlation do meet anti-monotonicity property crucial traditional methods. In paper, we heuristics vertical decomposition a database, pruning unproductive itemsets, and traversing set-enumeration tree itemsets tailored calculation N most significant rules, where can be specified by user. experimentally compared combination these three with previous approach. Our tests confirmed comutational performance improves several orders magnitude.