High Utility Itemset Mining using Path Encoded Utility Lists and Pattern Trees

Contenu principal de l'article

Vamsinath Javangula, Vani K. Suvarna, Haritha Dasari

Résumé

In this paper a high utility items mining algorithm mining items in a single phase has been proposed. The algorithm employs a novel data structure – path encoded utility lists for storing the utility, remaining utility and path information of the itemset. The utility is used for deciding whether a given itemset is high utility or not. The remaining utility is used for creating an upper bound which helps in deciding whether the projection of the itemset to its superset will result in high utility items. The path information guides in picking up the next itemset for generating the superset of the current itemset. The path encoded utility list is embedded in the node of a pattern tree, there by facilitating easy traversal of the problem space. Comparison with other related systems in the literature clearly shows the optimal performance of the algorithm in mining high utility itemsets.

Renseignements sur l'article

Rubrique
Articles