Citation Context

...interval to the left of 1/2, no incomparability but still some degree of indifference). Based on this preference structure, predictions can be derived using Algorithm 1 (or the decision rule given in =-=[21]-=- and [22]). As future work, we plan to elaborate on this alternative more closely. In particular, we are interested in the idea of adapting the overlap and grouping functions to the problem at hand, i...

Citation Context

...ing different information such as that coming from expert knowledge and information coming from mathematical models or empiric measures. New FRBCS models have been proposed on standard classification =-=[15]-=-, [16] and data streams [17] among others. They have also been also applied widely including, but not limited to, the detection of intrusions, medical applications and in the imbalanced data framework...

Citation Context

... from expert knowledge and information from mathematical models or empirical measures. Among the applications of FRBCSs we can find proposals in a variety of fields, including standard classification =-=[3]-=-, [4], detection of intrusions [5] or medical applications [6]. One goal of classification algorithms is to form a generalization from a set of labeled training instances so that classification accura...

Citation Context

...LASSIFICATION OF IMPRECISE DATA FISHER DISCRIMINATOR 721 where m1 and m2 are means of class c1 and c2, respectively, in input space. Betweenclass scatter matrix is defined as SB = (m2 − m1)(m2 − m1)T =-=(3)-=- So, the numerator of fraction is wT SBw. For simplification of denominator, it can be written as s̃21 = ∑ yi :xi∈R1 (yi − m̃1)2 = ∑ xi∈R1 (wT xi − wT m1)2 (4) = wT ⎛ ⎝ ∑ xi∈R1 (xi − m1)(xi − m1)T ⎞ ⎠...

Citation Context

...ming from mathematical models or empiric measures. There is a vast literature in the field of FRBCSs, which is very active at this time. New FRBCS models have been proposed on standard classification =-=[10]-=-, [11], [17] and data streams [2] among others. They have also been also applied widely including, but not limited to, the detection of intrusions [20], medical applications [1], [21], [19] and in the...

Citation Context

... highest nondominance degree is the predicted one Class = arg max i=1,...,m { 1− max j=1,...,m r′j i } (11) where R′ is the strict score-matrix. 4) Learning Valued Preference for Classification [33], =-=[34]-=-. This aggregation strategy, as in ND, considers the scorematrix as a fuzzy preference relation. In this manner, the original relation is decomposed into three new relations with different meanings: t...

Citation Context

...g was established by Hüllermeier and Brinker in [8], and was later employed by Hühn and Hüllermeier on a Fuzzy Round Robin Ripper approach based on a decision rule using fuzzy preference relations =-=[7]-=-. We will analyse the effect of this multiclassifier approach over a well-performing FRBCSs, the Fuzzy Hybrid Genetics-Based Machine Learning (FH-GBML) algorithm [12], in order to determine whether ou...

Citation Context

...en developed, for instance, new approaches in the framework of probability estimates [76], binary-tree based strategies [23], dynamic classification schemes [41] or methods using preference relations =-=[44,24]-=-, in addition to more classical well-known combinations such as Pairwise Coupling [39], Max-Wins rule [29] or Weighted Voting (whose robustness has been recently proved in [46]). In the specialized li...

Citation Context

...te values xi satisfy all the predicates in rA. Due to reasons of space we refrain from a detailed review of the RIPPER rule induction algorithm and instead refer to the extended version of this paper =-=[HH08]-=-. Round robin learning aka all-pairs or allversus-all learning is a special decomposition technique that transforms a multi-class classification problem involving m > 2 classes L = {λ1 . . . λm} into ...