Lavrac, N., Flach, P., Zupan, B.: Rule Evaluation Measures: A Unifying View. In: 9th Int. Workshop on Inductive Logic Programming. LNCS, Springer (1999) 17  Home/Search   Document Details and Download   Summary   Related Articles   Check

....to a small subset of the data. There are a number of ways in which the work here can be extended. Some of these are: 1. Other performance measures . We have concentrated on selection based on accuracy. ILP practitioners may be interested in other measures of performance (for example, see [11]) Ranking and selection procedures have been developed for distributions other than the Binomial [3, 6] It is clearly of interest to determine the extent to which these results can be used in ILP. 2. Selection based on screening . The indi erence zone approach is only one possible formulation ....

N. Lavrac, P. Flach, B. Zupan. Rule Evaluation Measures: A Unifying View. In S. Dzeroski and P.A. Flach, editors, Proceedings of the Ninth International Workshop on Inductive Logic Programming (ILP99), volume 1634 of LNAI, pages 174-185, Berlin, 1999. Springer.

....q 1 = R . S1 (B) and q 2 = R . S2 (B) are queries of C. However, as we shall see in Section 2.3, the con dence of such rules would not be the support of q 1 q 2 devided by the support of q 1 . 2. 3 Support and Con dence All standard measures of interest for association rules as given in [12], can also be de ned in our approach. For the purposes of this paper, however, we are only interested in the support and con dence of a rule. De nition 3 Support of a Query. Let C = hR; B; i be a context and q be a query such that sch(R) sch(q) For every instance I, the support of q ....

N. Lavrac, P. Flach et B. Zupan (1999). Rule Evaluation Measures: a Unifying View. Proc. of ILP'99.

No context found.

Lavra#c, N., Flach, P. and Zupan, B. #1999# Rule Evaluation Measures: A Unifying View. In Proceedings of the Ninth International Workshop on Inductive Logic Programming, volume 1634 of Lecture Notes in Arti#cial Intelligence: 74#

No context found.

Lavra c, N., Flach, P., & Zupan, B. (1999). Rule evaluation measures: A unifying view. Proceedings of the 9th International Workshop on Inductive Logic Programming (ILP-99) (pp. 174--185). Springer-Verlag.

No context found.

Lavra c, N., Flach, P., & Zupan, B. (1999). Rule Evaluation Measures: A Unifying View. In S. D zeroski & P. Flach (Eds.), Proceedings of the 9th International Workshop on Inductive Logic Programming (pp. 174--185). Springer-Verlag.

No context found.

N. Lavra c, P.A. Flach, and B. Zupan, `Rule evaluation measures: A unifying view', in Proceedings of the Nineth International Workshop on Inductive Logic Programming, pp. 74--185, Springer, (1999).

No context found.

N. Lavrac, P. Flach, and B. Zupan. Rule evaluation measures: A unifying view. In: S. Dzeroski and P. Flach (eds.): Proceedings of the 9th International Workshop on Inductive Logic Programming, 174--185, Springer-Verlag, 1999.

....for the number of examples of class H, and n(H.B) stand for the number of correctly classified examples (true positives) We use p(H.B) etc. for the corresponding probabilities. We then have that rule accuracy can be expressed as Acc(H B) p(H B) p(H. B) p(B) Weighted relative accuracy [14, 21], a reformulation of one of the heuristics used in MIDOS [22] is defined as follows. B) p(B) p(H B) p(H) 1) Weighted relative accuracy consists of two components: generality p(B) and relative accuracy p(H B) p(H) The second term, relative accuracy, is the accuracy gain relative ....

N. Lavrac, P. Flach, and B. Zupan. Rule evaluation measures: A unifying view. In Proc. of the 9th International Workshop on Inductive Logic Programming, 74--185. Springer, 1999.

....rules is used. The second one (CN2 wracc) uses weighted relative accuracy. The performance of two versions of CN2 was compared in sense of classification accuracy as well as number and length of the induced rules. 1 Introduction In the recent study of different rule evaluation measures [3], a new measure, named weighted relative accuracy (WRAcc) was proposed. It is proved to be equivalent to the novelty measure used in a descriptive induction framework. In contrast with widely accepted accuracy measure, WRAcc takes into account the improvement of the accuracy relative to the ....

.... rule if cond then c is equal to the conditional probability of class c, given that the condition cond is satisfied: Acc(if cond then c) p(cjcond) In the new version of CN2 (CN2 wracc) we replaced the accuracy measure by the weighted relative accuracy measure, recently introduced in [3]: WRAcc(if cond then c) p(cond) p(cjcond) Gamma p(c) 1) where p(cond) is the generality of the rule and p(c) is the prior probability of class c. The second term in Equation 1 is the accuracy gain relative to the default rule if true then c. The bigger the difference with the default ....

Lavrac, N., Flach, P. and Zupan, B. (1999) Rule Evaluation Measures: A Unifying View. In Proceedings of the Ninth International Workshop on Inductive Logic Programming, volume 1634 of Lecture Notes in Artificial Intelligence: 74--185. Springer-Verlag.

..... It is easy to show that this is equal to pHB Gamma HB , and thus symmetric in H and B. On the other hand, this measure has a certain interest: e.g. under simple expected frequency it equals p(B) p(H jB) Gamma p(H) which has been called weighted relative accuracy in a classification context [14]. Here, we call it the novelty associated with the rule, because it measures the novel information expressed by the rule that cannot be inferred from the evidence and the null hypothesis alone. Definition 3. The novelty of a rule 8(H B) is defined as Delta HB = HB Gamma p HB . Note that ....

N. Lavrac, P. Flach, and B. Zupan. Rule evaluation measures: A unifying view. In S. Dzeroski and P. Flach, editors, Proceedings of the 9th International Workshop on Inductive Logic Programming, volume 1634 of Lecture Notes in Artificial Intelligence, pages 174--185. SpringerVerlag, 1999.

....Stefan Institute Jamova 39, 1000 Ljubljana, Slovenia Ljupco.Todorovski ijs.si, Nada.Lavrac ijs.si 2 Department of Computer Science, University of Bristol The Merchant Venturers Building, Woodland Road, Bristol, UK Peter.Flach cs.bris.ac.uk Abstract. Weighted relative accuracy was proposed in [4] as an alternative to classification accuracy typically used in inductive rule learners. Weighted relative accuracy takes into account the improvement of the accuracy relative to the default rule (i.e. the rule stating that the same class should be assigned to all examples) and also ....

....dramatically reduces the size of the rule sets induced with CN2 (on average by a factor 9 on the 23 datasets we used) at the expense of only a small average drop in classification accuracy. 1 Introduction In a recent study of different rule evaluation measures in inductive machine learning [4], a new measure, named weighted relative accuracy (WRAcc) was proposed. This measure can be understood from different perspectives. On one hand, WRAcc takes into account the improvement of the accuracy relative to the default rule (i.e. the rule stating that the same class should be assigned to ....

Lavrac, N., Flach, P. and Zupan, B. (1999) Rule Evaluation Measures: A Unifying View. In Proceedings of the Ninth International Workshop on Inductive Logic Programming, volume 1634 of Lecture Notes in Artificial Intelligence: 74--

No context found.

Lavrac, N., Flach, P., Zupan, B.: Rule Evaluation Measures: A Unifying View. In: 9th Int. Workshop on Inductive Logic Programming. LNCS, Springer (1999) 17

No context found.

N. Lavrac, P. Flach, and B. Zupan. Rule Evaluation Measures: A Unifying View. In 9th International Workshop on Inductive Logic Programming, Lecture Notes in Computer Science. Springer, 1999.

No context found.

N. Lavrac, P. Flach, and B. Zupan. Rule Evaluation Measures: A Unifying View. In 9th International Workshop on Inductive Logic Programming, Lecture Notes in Computer Science. Springer, 1999.

No context found.

Lavrac, N., Flach, P., and Zupan, B. 1999. Rule Evaluation Measures: A Unifying View. In Proceedings of the 9th International Workshop on Inductive Logic Programming. Lecture Notes in AI, v. 1634, pp. 174-185.

No context found.

Lavrac, N., Flach, P., and Zupan, B. (1999) "Rule Evaluation Measures: A Unifying View". In: Proceedings of the 9th International Workshop on Inductive Logic Programming. Lecture Notes in AI, v. 1634, pp. 174-185.

No context found.

Lavrac, N., Flach, P., Zupan, B.: Rule Evaluation Measures: A Unifying View. In: 9th Int. Workshop on Inductive Logic Programming. LNCS, Springer (1999) 17

No context found.

N. Lavrac, P. Flach, and B. Zupan. Rule Evaluation Measures: A Unifying View. In 9th International Workshop on Inductive Logic Programming, Lecture Notes in Computer Science. Springer, 1999.

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