Catálogo de publicaciones - libros

In this contribution we carry out an analysis of the rule weights and Fuzzy Reasoning Methods for Fuzzy Rule Based Classification Systems in the framework of imbalanced data-sets with a high imbalance degree. We analyze the behaviour of the Fuzzy Rule Based Classification Systems searching for the best configuration of rule weight and Fuzzy Reasoning Method also studying the cooperation of some pre-processing methods of instances. To do so we use a simple rule base obtained with the Chi (and co-authors’) method that extends the well-known Wang and Mendel method to classification problems. The results obtained show the necessity to apply an instance pre-processing step and the clear differences in the use of the rule weight and Fuzzy Reasoning Method. Finally, it is empirically proved that there is a superior performance of Fuzzy Rule Based Classification Systems compared to the 1-NN and C4.5 classifiers in the framework of highly imbalanced data-sets.

This article deals with the identification of hinging hyperplane models. This type of non-linear black-box models is relatively new, and its identification is not thoroughly examined and discussed so far. They can be an alternative to artificial neural nets but there is a clear need for an effective identification method. This paper presents a new identification technique for that purpose based on a fuzzy clustering technique called Fuzzy c-Regression Clustering. To use this clustering procedure for the identification of hinging hyperplanes there is a need to handle restrictions about the relative location of the hyperplanes: they should intersect each other in the operating regime covered by the data points. The proposed method recursively identifies a hinging hyperplane model that contains two linear submodels by partitioning of the operating region of one local linear model resuling in a binary regression tree. Hence, this paper proposes a new algorithm for the identification of tree structured piecewise linear models, where the branches correspond to linear division of the operating regime based on the intersection of two local linear models. The effectiveness of the proposed model is demonstrated by a dynamic model identification example.

A vast majority of classification problems are characterized by an intrinsic vagueness in the knowledge of the class label associated to each example. In this paper we propose a classifier based on fuzzy discrete support vector machines, that takes as input a binary classification problem together with a membership value for each example, and derives an optimal separation rule by solving a mixed-integer optimization problem. We consider different methods for computing the membership function: some are based on a metric defined in the attribute space; some derive the membership function from a scoring generated by a probabilistic classifier; others make use of frequency voting by an ensemble classifier. To evaluate the overall accuracy of the fuzzy discrete SVM, and to investigate the effect of the alternative membership functions, computational tests have been performed on benchmark datasets. They show that fuzzy discrete SVM is an accurate classification method capable to generate robust rules and to smooth out the effect of outliers.

Different clustering algorithms are based on different similarity or distance measures (e.g. Euclidian distance, Minkowsky distance, Jackard coefficient, etc.). Jarvis-Patrick clustering method utilizes the number of the common neighbors of the k -nearest neighbors of objects to disclose the clusters. The main drawback of this algorithm is that its parameters determine a too crisp cutting criterion, hence it is difficult to determine a good parameter set. In this paper we give an extension of the similarity measure of the Jarvis-Patrick algorithm. This extension is carried out in the following two ways: (i) fuzzyfication of one of the parameters, and (ii) spreading of the scope of the other parameter. The suggested fuzzy similarity measure can be applied in various forms, in different clustering and visualization techniques (e.g. hierarchical clustering, MDS, VAT). In this paper we give some application examples to illustrate the efficiency of the use of the proposed fuzzy similarity measure in clustering. These examples show that the proposed fuzzy similarity measure based clustering techniques are able to detect clusters with different sizes, shapes and densities. It is also shown that the outliers are also detectable by the proposed measure.

In the paper we consider the problem of automatic fuzzy rules mining. A new method for generation of fuzzy rules according to the set of precedents is suggested. The proposed algorithm can find all significant rules with respect to wide range of reasonable criterion functions. We present the statistical criterion for knowledge quality estimation that provides high generalization ability. The theoretical results are complemented with the experimental evaluation.

The paper proposes three fuzzy regression models - concerning temperature and electricity load - based on real data. In the first two models the monthly temperature in a period of four years in a Polish city is analyzed. We assume the temperature to be fuzzy and its dependence on time and on the temperature in the previous month is determined. In the construction of the fuzzy regression models the least square methods was used. In the third model we analyze the dependence of the daily electricity load (assumed to be a fuzzy number) on the (crisp) temperature. Outliers, i.e. non-typical instances in the observations are identified, using a modification of an identification method known from the literature. The proposed method turns out to identify the outliers consistently with the real meaning of the experimental data.

In this paper we propose the Possibilistic C -Means in Feature Space and the One-Cluster Possibilistic C -Means in Feature Space algorithms which are kernel methods for clustering in feature space based on the ossibilistic approach to clustering. The proposed algorithms retain the properties of the possibilistic clustering, working as density estimators in feature space and showing high robustness to outliers, and in addition are able to model densities in the data space in a non-parametric way. One-Cluster Possibilistic C -Means in Feature Space can be seen also as a generalization of One-Class SVM.

OpenAdap.net is an Open Source project aimed at breaking the barriers existing in the flow of information access and information processing. The infrastructure makes it possible the dissemination of resources like knowledge, tools or data, their exposure to evaluation in ways that might be unanticipated and hence support the evolution of communities of users around a specific domain. The architecture is designed by analogy with a virtual distributed operating system in which the dynamic resources are presented as files in a structured virtual file system featuring ownership and access permissions.

The possibility of solving an optimization problem by an exhaustive search on all the possible solutions can advantageously replace traditional algorithms for learning neuro-fuzzy networks. For this purpose, the architecture of such networks should be tailored to the requirements of quantum processing. In particular, it is necessary to introduce superposition for pursuing parallelism and entanglement. In the present paper the specific case of neuro-fuzzy networks applied to binary classification is investigated. The peculiarity of the proposed method is the use of a nonlinear quantum algorithm for extracting the optimal neuro-fuzzy network. The computational complexity of the training process is considerably reduced with respect to the use of other classical approaches.

This paper presents an algorithm to compute high dimensional piecewise linear (PWL) functions with simplicial division of the input domain, and introduces the circuit scheme for its implementation in a FPGA. It is also investigated how to modify this algorithm and implementation to compute a class of PWL fuzzy systems.