Lent, B., Swami, A.N., Widom, J.: Clustering association rules. In: Proceedings of the 13th International Conference on Data Engineering (ICDE). (1997) 220--231  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Most e orts have focused on developing novel algorithms and data structures to aid ecient computation of such rules. Despite major e orts, the computational complexity of even the best known methods remains high. While several ecient algorithms for association rule mining have been proposed [1, 6, 20, 15, 9, 19, 14, 21], overall eciency is still a major issue, specially for other kinds of rule induction such as ratio rules [8] and chi square rules [4] While many forms of rule inductions are interesting, association rules were found to be appealing because of their simplicity and intuitiveness. In this ....

Brian Lent, Arun N. Swami, and Jennifer Widom. Clustering association rules. In Proc of the 3th ICDE, pages 220-231, 1997.

....In [LHM98] association rule mining is used to build classifiers by focusing on a special subset of association rules with consequent restricted to the class label attribute. This study shows that it is possible to build an accurate classifier for prediction from the set of generated rules. In [LSW97], an approach is proposed to perform clustering on association rules. It derives association rules first, then clusters all those two attribute association rules where the consequent of the rules satisfiers some segmentation criteria. This approach can be used for data segmentation. 2.13. Other ....

B. Lent, A. Swami, and J. Widom, "Clustering Association Rules," Proceedings of the IEEE International Conference on Data Engineering, Birmingham, England, April 1997, pp. 220-231.

.... of mining frequent patterns arose rst as a sub problem of mining association rules, but then it turned out that frequent patterns solve a variety of problems: mining sequential patterns [AS95] episodes [MTV97] association rules [AS94] correlations [BMS97, SBM98] multi dimensional patterns [KHC97, LSW97], maximal patterns [ZPOL97, LK98] and several other important knowledge discovery tasks [HPY00] Since the complexity of this problem is exponential in the size of the binary database input relation and since this relation has to be scanned several times during the process, ecient algorithms for ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. Proc. ICDE conf., pp 220-231, March 1997.

....topic. The problem of mining frequent patterns arose rst as a sub problem of mining association rules [1] but it then turned out to be present in a variety of problems [18] mining sequential patterns [3] episodes [26] association rules [2] correlations [10, 37] multi dimensional patterns [21, 22], maximal patterns [8, 53, 23] closed patterns [47, 31 33] Since the complexity of this problem is exponential in the size of the binary database input relation and since this relation has to be scanned several times during the process, ecient algorithms for mining frequent patterns are ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proc. of the 13th Int'l Conf. on Data Engineering (ICDE), pages 220-231, Mar. 1997.

....Classification and Association Rules by proposing adaptations to the CBA Algorithm Davy Janssens GeertWets Tom BrUs Koen Vanhoof DAVY.JANS SENSLUC.AC.BE GEERT.WETSLUC.AC.BE TOM.BRIJSLUC.AC.BE KOEN.VANHOOFLUC.AC. BE Limburg University Centre, Universitaire Campus, gebouw D, B 3590 Diepenbeek, Belgium Abstract In recent years, extensive research has been carded out by focusing on association rules to build more accurate classifiers. These integrated approaches mainly focus on a limited subset of association rules, i.e. those rules where the consequent of the rule is ....

....it with another measurement of the quality of association rules: i.e. intensity of implication. The new algorithm has been implemented and empirically tested on an authentic mancial dataset for purposes of bankruptcy prediction. We validated our results with an association ruleset, with C4. 5, with original CBA and with CART by statistically comparing its performance via the area under the ROC curve. The adapted CBA algorithm presented in this paper proved to generate significantly better results than the other classifiers at the 5 level of significance. 1. Introduction ....

[Article contains additional citation context not shown here]

Lent, B., Swami, A.N., Widom, J. (1997). Clustering association rules. In Proc. of the Thirteenth Intemational Conference on Data Engineering, Birmingham, U.K, p. 220-231

....represented as a FDT is understandable and it differs from black box systems as neural networks. Moreover, a FDT is equivalent to a set of fuzzy rules [6] And such kind of induced rules can be introduced to optimize the query process of the database [7, 33] or to deduce decisions from data [1, 2, 15, 16]. FDTs enable us to obtain various kinds of such rules [19] Thus, it is a powerful knowledge representation. A FDT, as a set of fuzzy rules, can be used as knowledge base to help flexibly querying a database. Nowadays, literature related to FDT construction is very active. However, few works ....

....of knowledge that can be associated with the database. This knowledge can be used in different ways. First of all, it enables us to improve the querying process of the database. A set of rules can be introduced to optimize the query process of the database [7, 33] or to deduce decisions from data [1, 2, 15, 16]. Such a set of induced fuzzy rules can be associated with a database as a knowledge base that can be used to help answering frequent queries. A fast response can be found for a query on the value of an attribute. It can also lower the conditions on the values of attributes for a query, before the ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proceedinds of the 13th International Conference on Data Engineering, pages 220 231, Birmingham, UK, April 1997. IEEE Computer Society Press.

....A particular instance of such knowledge is represented by fuzzy knowledge. We study induction of a set of fuzzy rules from a database, the inductive learning scheme. Such kind of induced rules can be introduced to optimize the query process of the database [8, 16] or to deduce decisions from data [2, 3, 11, 12]. Many works have been done on the topics of inducing rules from database, but the introduction of fuzzy set theory in the process of induction of rules is more recent. In the induction of knowledge from data, difficulties appear when considering data described by means of numeric symbolic ....

....from a database [2, 3] In this case, there is no particular highlighted attribute from .4. The aim is to find rules that describe the examples of in pointing out association rules between values of attributes of .4. Association rules can be considered as a particular form of classification rules [12]. To induce a set of such rules from a training set, several trees can be generated in varying the attribute selected as the class. The values of attributes appearing on a path of the decision tree can be considered as linked to determine the class. In the tree, each path is associated with a ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proceedinds of the 13th International Conference on Data Engineering, pages 220 231, Birmingham, UK, April 1997. IEEE Computer Society Press.

....the previous section as L(f) s(# y ) s(# f(x) 2s(# y # f(x) or, with other words, misclassification occurs if either y is true and f(x) is not true or f(x) is true and y is not. Association rule technology has been applied to to classification and examples of this approach are in [28, 52, 51, 53, 57]. 4.2.3 Function spaces We will consider several classes of functions. The simplest class is the class of constant functions f(x) f 0 . More complex classes are obtained by exploiting various function properties and properties of Y and X. In a first class we will partition X using the tensor ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proc. 1997.

....rule(s) from among all the rules mined for classification. Since association rules explore highly confident associations among multiple variables, it may overcome some constraints introduced by a decision tree induction method which examines one variable at a time. Extensive performance studies [6, 9, 3, 11] show that associationbased classification may have better accuracy in general. However, this approach may also suffer some weakness as shown below. On one hand, it is not easy to identify the most effective rule at classifying a new case. Some method, such as [9, 3, 11] simply selects a rule ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. In ICDE'97, England, April 1997.

.... based on the levelwise Apriori framework [3, 13] partitioning [19, 18] and sampling [24] ii) incremental updating and parallel algorithms [6, 2, 8] iii) mining of generalized and multi level rules [21, 9] iv) mining of quantitative rules [22, 16] v) mining of multidimensional rules [7, 14, 12]; vi) mining rules with item constraints [23] and (vii) association rule based query languages [15, 4] However, from the standpoint of the user s interaction with the system, the process of association mining can be summarized as follows. First, the user specifies the part of the database to be ....

....Phase II, the user can specify the desired significance metric, and can give different conditions that must be satisfied by the antecedent and consequent of the relationships to be formed. There are already several proposals in the literature that make the notion of associations less rigid [5, 7, 9, 12, 14, 21]. We are not proposing another here. Instead, we are proposing an architecture that allows many of those alternative notions to co exist, and that permits the user to choose whatever is appropriate for the application. 2 Architecture Figure 1 shows a two phase architecture for exploratory ....

B. Lent, A. Swami, and J. Widom. Clustering association rules. ICDE 97.

....discovered rules into classes, according to some similarity criteria, that biologists could subsequently explore and analyze. In this section, we present a similarity based rule grouping method for this purpose. The problem of clustering similar rules was first studied in the KDD community in [18]. In particular, Lent et al. [18] clusters discovered association rules in the two dimensional space using heuristic methods based on geometric properties of twodimensional grids. This approach is restricted only to the rules that have two fixed attributes (both of them discrete and ordered) in ....

....according to some similarity criteria, that biologists could subsequently explore and analyze. In this section, we present a similarity based rule grouping method for this purpose. The problem of clustering similar rules was first studied in the KDD community in [18] In particular, Lent et al. [18] clusters discovered association rules in the two dimensional space using heuristic methods based on geometric properties of twodimensional grids. This approach is restricted only to the rules that have two fixed attributes (both of them discrete and ordered) in their antecedents, although some ....

[Article contains additional citation context not shown here]

Lent, B, Swami, A.N. and Widom, J. Clustering association rules. In Proceedings of International Conference on Data Engineering, 1997.

....association rules can be applied to these items. Unfortunately, the Max support threshold may exclude some strong and interesting rules from being discovered. In order to overcome this problem, in our model we will use a more meaningful metric, namely density, to restrict the search space. [6] presented a geometric based algorithm, BitOp, for performing clustering on association rules of the form A B = C where the left hand side attributes (A and B) are quantitative and the right hand side attribute (C) is categorical. Given a rule format A B = C, the attribute domain is ....

....rules , mining association rules with constraints, mining sequential patterns in time series data, discovering periodic patterns, and similarity search in time sequence(s) Interested readers please refer to [13] for a detailed overview. Alternative solutions to our problem Inspired by [9] and [6], the following are two alternative solutions to our problem. SR algorithm: One method is to map the numerical attribute evolutions to binary attributes and apply one of the existing algorithms for mining association rules [9] For each numerical attribute A, its domain is quantized to b ....

[Article contains additional citation context not shown here]

B. Lent, A. Swami, and J. Widom. Clustering association rules. Proc. 13th Int. Conf. on Data Engineering, 220-231, 1997.

....for storing aggregated rules, the grouping algorithm is linear in the total size of the rules to be grouped. This is especially important in applications dealing with very large numbers of rules. 3. 5 Related Work There have been related approaches to rule grouping proposed in the literature [51, 86] that consider association rules in which both numeric and categorical at tributes can appear in the body and only categorical attributes in the head of a rule. However, 51] take a more restrictive approach by allowing only two numeric at tributes in the body and one categorical attribute in ....

....numbers of rules. 3.5 Related Work There have been related approaches to rule grouping proposed in the literature [51, 86] that consider association rules in which both numeric and categorical at tributes can appear in the body and only categorical attributes in the head of a rule. However, [51] take a more restrictive approach by allowing only two numeric at tributes in the body and one categorical attribute in the head of a rule, whereas [86] 43 allow any combination of numeric and categorical attributes in the body and one or more categorical attributes in the head of a rule. Both ....

[Article contains additional citation context not shown here]

B. Lent, A. N. Swami, and J. Widom. Clustering association rules. In Proceed- ings of the Thirteenth International Conference on Data Engineering, April 7-11, 1997.

No context found.

Lent, B., Swami, A.N., Widom, J.: Clustering association rules. In: Proceedings of the 13th International Conference on Data Engineering (ICDE). (1997) 220--231

No context found.

Lent, B., Swami, A., & Widom, J. (1997). Clustering association rules. Proc. of ICDE.

No context found.

Brian Lent, Arun N. Swami, and Jennifer Widom. Clustering association rules. In ICDE, pages 220--231, 1997.

No context found.

B. Lent, et al. Clustering association rules. ICDE'97.

No context found.

Lent, B., Swami, A. N., and Widom, J. 1997. Clustering association rules. In ICDE. 220--231.

No context found.

Brian Lent, Arun N. Swami, and Jennifer Widom. Clustering association rules. In ICDE, pages 220--231, 1997.

No context found.

Brian Lent, Arun N. Swami, and Jennifer Widom. Clustering association rules. In Proceedings of the 13th International Conference on Data Engineering, pages 220--231. IEEE Computer Society, 1997.

No context found.

Brian Lent, Arun N. Swami, and Jennifer Widom. Clustering association rules. In Proc of the 3th ICDE, pages 220-231, 1997.

No context found.

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proc. 1997.

No context found.

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proc. 1997 Int. Conf. Data Engineering (ICDE'97), pages 220--231, Birmingham, England, April 1997.

No context found.

B. Lent, A. Swami, and J. Widom. Clustering association rules. In Proceedings of the Thirteenth International Conference on Data Engineering, pp 220--231. Birmingham, UK, April 1997.

No context found.

Lent, B., Swami, A., & Widom, J. (1997). Clustering association rules. Proc. of ICDE.

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC