A Novel Rule Weighting Approach in Classification Association Rule Mining

Abstract

Classification Association Rule Mining (CARM) is a recent Classification

Dynamic K: A Novel Satisfaction Mechanism for CAR-Based Classifiers

“…In Table 5, the results show that "Dynamic K" yields an average accuracy higher than all other evaluated mechanisms, having a difference of 0.92% with respect to the mechanism in the second place ("Best K rules" with K set to 5, the same value used in other works [5,11,12,14]). Additionally, "Dynamic K" wins in 16 of the 20 datasets and ties in the other four.…”

“…As we mentioned above, the three satisfaction mechanisms reported have limitations that can affect the classification accuracy. In general, the "Best K Rules" mechanism has been the most widely used for CAR-based classifiers, reporting the best results [11]. However, using this mechanism could affect the classification accuracy.…”

Studying Netconf in Hybrid Rule Ordering Strategies for Associative Classification

“…2. Best K Rules: For each pre-defined class, this mechanism selects the first (top) K rules satisfying the transaction to be classified and it assigns the class for this transaction applying an averaging process as those used in [20]. 3.…”

“…In this paper, we propose four novel hybrid rule ordering strategies, which combine the Netconf measure with four rule ordering strategies based on Support and Confidence, as well as with a rule ordering strategy based on the CAR's size. Our experiments were conducted using several datasets taken from the UCI Machine Learning Repository [3], all of them used in other representative works [20][21][22]. The results obtained using the proposed hybrid rule ordering strategies show good performance, regarding the accuracy of classification, compared against the best hybrid rule ordering strategies reported in the literature.…”

Association rule with frequent pattern growth algorithm for frequent item sets mining

“…The experimental results show that the proposed algorithm has better classification accuracy in comparison with classification association rule mining. Wang et al, [2] presented a new rule weighting techniques in association rule mining called class item score based rule weighting (CISRW) algorithm. The experimental results indicate that the CISRW algorithm based on related rule ordering techniques do well by means of accuracy of classification.…”