Carma is a type of online association algorithm, designed to facilitate association rule with online data flow and successively changing support thresholds. In this paper we study the factors that contribute to the efficiency of Carma and how data flow distribution give effects on the performance of Carma. We design several experiments with two kinds of data. In fixed support threshold situations, we compare Carma with that of Apriori. We find the sets generated by Carma are subsets of those generated by Apriori. We find that if the support threshold is reasonably defined, these two algorithms reach the same results. On the other hand, as the support threshold increases, Phase I generates less items and the number of deleted sets from Phase II first increases and then declines. Carma behaves consistently towards changing support. We notice the earlier the items enter into a lattice, the more accurate the estimations are. If base stone elements show up early in the transaction, the performance of Phase II is mainly influenced by the late-entered item sets. Based on the discussion with Carma, we propose a new procedure to improve Carma. Simulations reveal that the modified algorithm works well.
|主出版物標題||NCM 2009 - 5th International Joint Conference on INC, IMS, and IDC|
|出版狀態||已發佈 - 2009|
|事件||NCM 2009 - 5th International Joint Conference on Int. Conf. on Networked Computing, Int. Conf. on Advanced Information Management and Service, and Int. Conf. on Digital Content, Multimedia Technology and its Applications - Seoul, 大韓民國|
持續時間: 8月 25 2009 → 8月 27 2009
|其他||NCM 2009 - 5th International Joint Conference on Int. Conf. on Networked Computing, Int. Conf. on Advanced Information Management and Service, and Int. Conf. on Digital Content, Multimedia Technology and its Applications|
|期間||8/25/09 → 8/27/09|
ASJC Scopus subject areas