International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems

Feature Selection is an important task within Text Categorization, where irrelevant or noisy features are usually present, causing a lost in the performance of the classifiers. Feature Selection in Text Categorization has usually been performed using a filtering approach based on selecting the features with highest score according to certain measures. Measures of this kind come from the Information Retrieval, Information Theory and Machine Learning fields. However, wrapper approaches are known to perform better in Feature Selection than filtering approaches, although they are time-consuming and sometimes infeasible, especially in text domains. However a wrapper that explores a reduced number of feature subsets and that uses a fast method as evaluation function could overcome these difficulties. The wrapper presented in this paper satisfies these properties. Since exploring a reduced number of subsets could result in less promising subsets, a hybrid approach, that combines the wrapper method and some scoring measures, allows to explore more promising feature subsets. A comparison among some scoring measures, the wrapper method and the hybrid approach is performed. The results reveal that the hybrid approach outperforms both the wrapper approach and the scoring measures, particularly for corpora whose features are less scattered over the categories.

The aim of this paper is to develop a fuzzy linear programming technique for multidimensional analysis of preference (FLINMAP) in multiattribute group decision making problems with linguistic variables and incomplete preference information. In this paper, linguistic variables are used to assess an alternative on qualitative attributes using fuzzy ratings corresponding to some triangular fuzzy numbers. Each alternative is assessed on the basis of its distance to a fuzzy positive ideal solution (FPIS) which is unknown a priori. The FPIS and the weights of attributes are calculated by constructing a new linear programming model based on the group consistency and inconsistency indices defined on the basis of preferences between alternatives given by the decision makers. The distance of each alternative to the FPIS can be calculated to determine the ranking order of all alternatives. The implementation process of this methodology is demonstrated with an example.

We will discuss some relations among values, i.e., global / overall importance of each single element of the object to be considered, interaction indices, i.e., simultaneous interaction among the elements in a coalition, and inclusion-exclusion coverings, i.e., decomposability of non-additive measures and / or of the Choquet integral. So, we show that inclusion-exclusion coverings are characterized by the Shapley interaction index. Then, inclusion-exclusion coverings are invariant under the transformations which characterize the Shapley value and / or the Shapley interaction index, and preserve the Shapley value.

This paper describes an approach to pointwise construction of general aggregation operators, based on monotone Lipschitz approximation. The aggregation operators are constructed from a set of desired values at certain points, or from empirically collected data. It establishes tight upper and lower bounds on Lipschitz aggregation operators with a number of different properties, as well as the optimal aggregation operator, consistent with the given values. We consider conjunctive, disjunctive and idempotent n-ary aggregation operators; p-stable aggregation operators; various choices of the neutral element and annihilator; diagonal, opposite diagonal and marginal sections; bipolar and double aggregation operators. In all cases we provide either explicit formulas or deterministic numerical procedures to determine the bounds. The findings of this paper are useful for construction of aggregation operators with specified properties, especially using interpolation schemata.

Information Retrieval Systems (IRSs) based on an ordinal fuzzy linguistic approach present some problems of loss of information and lack of precision when working with discrete linguistic expression domains or when applying approximation operations in the symbolic aggregation methods. In this paper, we present a new IRS model based on the 2-tuple fuzzy linguistic approach, which allows us to overcome the problems of ordinal fuzzy linguistic IRSs and improve their performance.

This section is maintained by Vladik Kreinovich. Please send your abstracts (or copies of papers that you want to see reviewed here) to vladik@utep.edu, or by regular mail to: V. Kreinovich, Department of Computer Science, University of Texas, El Paso, TX 79968, USA.