Gediga, G. and Duntsch, I. Rough approximation quality revisited, Artificial Intelligence, 132, 219-234, 2001. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Mining High Order Decision Rules - Yao (Correct)
....one often uses an attribute value language to represent individual objects and mined knowledge [9 11,13,14,20] Each object is represented by the values of a set of attributes. The knowledge mined from a dataset is represented in the form of rules. At least two types of rules can be identified [5]. The first type, called type 1 rules in this paper, is exemplified by decision rules. A type 1 rule states that if the value of an object is v a on attribute a, then the value of the object is v b on attribute b . By pooling together many type 1 rules, we can obtain the second type rules, called ....
....better 5 small 2 years 200 very light good Size : small Size middle Size large, 3 years 2 years, 200 250 300, very light light heavy very heavy, best better good. The order relations induces the following orderings of products: Size : [3, 4, 5] # Size [1] # Size [2] 1, 2, 3, 4] # Warranty [5] Price : 1, 5] # Price [4] # Price [2, 3] 4, 5] # Weight [2] 1] # Overall [4] # Overall [2, 3, 5] Examples of formulas and their meaning sets are given by: m(# Size ) 2) 3, 1) 3, 2) ....
[Article contains additional citation context not shown here]
Gediga, G. and Duntsch, I. Rough approximation quality revisited, Artificial Intelligence, 132, 219-234, 2001.
On Model Evaluation, Indices of Importance, and Interaction.. - Gediga, Düntsch (2002) Self-citation (Gediga Dntsch) (Correct)
....of the g index by assuming that the transformation somehow fits to the semantics (the significance or importance ) We have shown that d 1 and d 2 are not necessarily meaningful, either for admissibility or usefulness, due to a lack of a sound theory which guides the index building process. In [10] we have introduced the PRE (Proportional Reduction of Errors) approach of Hildebrand et al. 15] into RSDA, which in the general case describes the error reduction when a model is applied, based on the errors of a given benchmark model. In the context of RSDA, we say that an error is an ....
....sets, which are constructed by random assignment of elements to the attributes S. Set of attribute sets, which are constructed by random assignment of elements to the attribute a. Error of the model Error of the benchmark model Interpretation Source 0UWV 1 admissibility [10] S#XZY g R U [ usefulness [10] 1 g P aabUWV 1 aa U c 0 admissible gain this text SdX Y aa U c R usable gain [7] SUfV 1 SU c 0 admissible set gain this text S#X Y SU c R usable set gain this text ....
[Article contains additional citation context not shown here]
Gediga, G. and Dntsch, I. (2001). Rough approximation quality revisited. Artificial Intelligence, 132:219--234.
Approximation Quality for Sorting Rules - Gediga, Düntsch Self-citation (Dntsch Gediga) (Correct)
....There is, of course, an infinite number of possible schemes; on the basis of Occam s razor, we choose simple ones which seem to do what we want it to do. This is similar to the fact that the classical is only one of infinitely many measures which could be used, depending on the circumstances [9]. 4 A proposal for multi criteria sorting quality In order to make clearer where the problems lie, we recall the measure suggested by Greco et al. 13] for the approximation quality of rules of the form (8) Let us consider a partial order on ( and a linear order on ( We will ....
Dntsch, I. & Gediga, G. (2000). Rough approximation quality revisited. Preprint. 27
On Model Evaluation, Indices of Importance, and Interaction.. - Gediga, Düntsch Self-citation (Gediga Dntsch) (Correct)
....the g index by assuming that the transformation somehow fits to the semantics (the significance or importance ) We have shown that d 1 and d 2 are not necessarily meaningful, either for admissibility or usefulness, due to a lack of a sound theory which guides the index building process. In [10] we have introduced the PRE (Proportional Reduction of Errors) approach of Hildebrand et al. 15] into RSDA, which in the general case describes the error reduction when a model is applied, based on the errors of a given benchmark model. In the context of RSDA, we say that an error is an ....
....constructed by random assignment of elements to the attributes S. C s : Set of attribute sets, which are constructed by random assignment of elements to the attribute a. Error of the model Error of the benchmark model Interpretation Source 1 S g T P U 1 S g T 0UWV 1 admissibility [10] 1 S g T P U 1 S#XZY g R U [ R Q s usefulness [10] 1 S g T P U 1 g P aabUWV 1 S g T T P aa U c 0 U admissible gain this text 1 S g T P U 1 SdX Y g T T P 4 aa U c R Ue[ R C s ] usable gain [7] 1 S g T P U 1 S g T P SUfV 1 ....
[Article contains additional citation context not shown here]
Gediga, G. and Dntsch, I. (2001). Rough approximation quality revisited. Artificial Intelligence, 132:219--234.
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