CPAR: Classification based on Predictive Association Rules

Abstract

Recent studies in data mining have proposed a new classification approach, called associative classification, which, according to several reports, such as [7,6], achieves higher classification accuracy than traditional classification approaches such as C4.5. However, the approach also suffers from two major deficiencies: (1) it generates a very large number of association rules, which leads to high processing overhead; and (2) its confidence-based rule evaluation measure may lead to overfitting.In comparison with associative classification, traditional rule-based classifiers, such as C4.5, FOIL and RIPPER, are substantially faster but their accuracy, in most cases, may not be as high. In this paper, we propose a new classification approach, CPAR (Classification based on Predictive Association Rules), which combines the advantages of both associative classification and traditional rule-based classification. Instead of generating a large number of candidate rules as in associative classification, CPAR adopts a greedy algorithm to generate rules directly from training data. Moreover, CPAR generates and tests more rules than traditional rule-based classifiers to avoid missing important rules. To avoid overfitting, CPAR uses expected accuracy to evaluate each rule and uses the best k rules in prediction.

Feature Construction Based on Closedness Properties Is Not That Simple

“…4. Then, we have compared FC & C4.5 with other associative classification approaches, namely CBA [7], CMAR [14], CPAR [15], and an EPs-based classifier SJEP-classifier [26]. Accuracy results for associative classifiers are taken from [14].…”

“…For example, CBA [7] ranks the rules and it uses the best one to label x. Other algorithms choose the class that maximizes a defined score (CMAR [14] uses combined effect of subsets of rules when CPAR [15] uses average expected accuracy of the best k rules). Also, starting from ideas for class characterization [16], [17] is an in-depth formalization of all these approaches.…”

Arguing from Experience to Classifying Noisy Data

“…This algorithm was later further developed by Yin and Han to produce the PRM (Predictive Rule Mining) CAR generation algorithm PRM was then further developed, by Yin and Han, to produce CPAR (Classification based on Predictive Association Rules) [28].…”

Credit scoring using ensemble of various classifiers on reduced feature set

“…To improve the less accuracy of parametric statistical methods, many models based on data-mining methods are built. These methods include the decision trees (DT) (Daviset al, 1992), (Frydman et al, 1985), (Zhou and Zhang (2008)); artificial neural networks (ANN) (Jensen (1992)), (West (2000)), (West et al, 2005); k-nearest neighbour (Henley and Hand (1996)), genetic programming (GP) (Abdou (2009)), (Onget al, 2005); genetic algorithm (GA) (Desai (1997)), (Walker et al,1995), (Zhang et al, 2007); case-based reasoning (CBR) (Chuang and Lin (2009)), (Jo et al, 1997), (Park and Han (2002)); Artificial Immune System Algorithm (Leung et al, 2007); rule extraction based on NN (Setionoet al, 2008); classification based on association rules (Li et al, 2001), (Liu et al, 1998), (Yin and Han (2003)) and support vector machines (SVM) , , , etc.…”