Catálogo de publicaciones - libros

One of the concerns of knowledge discovery in databases is the production of association rules. An association rule defines a relationship between two sets of attributes and , caracterising the data studied. Such a rule means that objects sharing attributes of will “likely” have those contained in . Yet, this notion of “likeliness” depends on the datamining context.

Many interestingness measures have been introduced in order to quantify this likeliness. This panel of measures is heterogeneous and the ranking of extracted rules, according to measures, may differ largely.

This contribution explores a new approach for assessing the quality of rules: aggregating valued relations. For each measure, a valued relation is built out of the numerical values it takes on the rules, and represents the preference of a rule over another. The aim in using such tools is to take into account the intensity of preference expressed by various measures, and should reduce incomparability issues related to differences in their co-domains. It also has the advantage of relating the numerical nature of measures compared to pure binary approaches.

We studied several aggregation operators. In this contribution we discuss results obtained on a toy example using the simplest of them.

The OWA weighting vector and the fuzzy quantifiers are strictly related. An intuitive way for shaping a monotonic quantifier, is by means of the threshold that makes a separation between the regions of what is satisfactory and what is not. Therefore, the characteristics of a threshold can be directly related to the OWA weighting vector and to its metrics: the attitudinal character and the entropy (or dispersion). Generally these two metrics are supposed to be independent, although some limitations in their value come when they are considered jointly. In this paper we argue that these two metrics are strongly related by the definition of quantifier threshold, and we show how they can be used jointly to verify and validate a quantifier and its threshold.

Subjective nature of system components makes the natural problems more complicated and harder to quantify. Thus, effective reflection of uncertainties, which is essential for generating reliable and realistic outcomes, has become a major concern for risk assessment. With the growing trend of fuzzy modeling and simulation of environmental problem, there is a need to develop a risk analysis approach which can use the fuzzy number output for characterization of risk. This study has been done to fulfil that need. Integration of system simulation and risk analysis using fuzzy approach allowed to incorporate system modelling uncertainty and subjective risk criteria. In this study, an integrated fuzzy relation analysis (IFRA) model is proposed for risk assessment involving multiple criteria. Model is demonstrated for a multi-components groundwater contamination problem. Results reflect uncertainties presented as fuzzy number for different modelling inputs obtained from fuzzy system simulation. Integrated risk can be calculated at different membership level which is useful for comprehensively evaluating risks within an uncertain system containing many factors with complicated relationship. It has been shown that a broad integration of fuzzy system simulation and fuzzy risk analysis is possible.

This paper presents a new watermarking method for protecting non-numerical databases. The proposed watermarking system allows the data owner to define a similarity function in order to reduce the distortion caused by watermark embedding while, at the same time, reducing the number of element modifications needed by the embedding process. A mathematical analysis is provided to justify the robustness of the mark against different types of malicious attacks. The usefulness of this extensible and robust method is illustrated by describing some application domains and examples.

Schema and record matching are tools to integrate files or databases. Record linkage is one of the tools used to link those records that while belonging to different files correspond to the same individual.

Standard record linkage methods are applied when the records of both files are described using the same variables. One of the non-standard record linkage methods corresponds to the case when files are not described using the same variables.

In this paper we study record linkage for non common variables. In particular, we use a supervised approach based on neural networks. We use a neural network to find the relationships between variables. Then, we use these relationships to translate the information in the domain of one file into the domain of the other file.

Genomewide analysis of linkage disequilibrium (LD) is commonly based in the maximum likelihood estimator. This estimator of LD suffers of a well known bias toward disequilibrium that becomes particularly serious in small samples with SNPs that are not very common in the population. Algorithms able to identify LD patterns, such as haplotype blocks or LD decay maps do a non-random selection of SNPs to be included in the analysis in order to remove this bias. However, they introduce ascertainment bias that can mask the real decay of disequilibrium in the population, with several consequences on the validity and reproducibility of genetic studies. In this work, we use a new Bayesian estimator of LD that greatly reduces the effect of ascertainment bias in the inference of LD decay. We also provide a software that use the Bayesian estimator to compute pairwise LD from SNP samples.

Because of the importance of companies’ financial distress prediction, this paper applies support vector machine (SVM) to the early-warning of financial distress. Taking listed companies’ three-year data before special treatment (ST) as sample data, adopting cross-validation and grid-search technique to find SVM model’s good parameters, an empirical study is carried out. By comparing the experiment result of SVM with Fisher discriminant analysis, Logistic regression and back propagation neural networks (BP-NNs), it is concluded that financial distress early-warning model based on SVM obtains a better balance among fitting ability, generalization ability and model stability than the other models.

Verification of probabilistic systems is usually based on variants of Markov processes. For systems with continuous dynamics, Markov processes are generated using discrete approximation methods. These methods assume an exact model of the dynamic behavior. However, realistic systems operate in the presence of uncertainty and variability and they are described by uncertain models. In this paper, we address the problem of probabilistic verification of uncertain systems using Bounded-parameter Markov Decision Processes (BMDPs). Proposed by Givan, Leach and Dean [1], BMDPs are a generalization of MDPs that allow modeling uncertainty. In this paper, we first show how discrete approximation methods can be extended for modeling uncertain systems using BMDPs. Then, we focus on the problem of maximizing the probability of reaching a set of desirable states, we develop a iterative algorithm for probabilistic verification, and we present a detailed mathematical analysis of the convergence results. Finally, we use a robot path-finding application to demonstrate the approach.

This work is focus in the following critical questions: Our proposal is based in a specification language which has the following properties: a) it permits to give specifications of modalities in a natural way; b) it defines a process of argumentation reasoning considering modalities; and c) it permits to build arguments from incomplete information.

Feature selection is a well-researched problem, which can improve the network performance and speed up the training of the network. In this paper, we proposed an effective feature selection scheme for SVM using the hybrid of enhanced genetic algorithm and mutual information, in which mutual information between each input and each output of the data set is employed in mutation in evolutionary process to purposefully guide search direction based on some criterions. In order to avoid the noise fitness evaluation, in evaluating the fitness of an input subset, a SVM should adaptively adjust its parameters to obtain the best performance of network, so an enhanced GA is used to simultaneously evolve the input features and the parameters of SVM. By examining two real financial time series, the simulation of three different methods of feature selection shows that the feature selection using the hybrid of GA and MI can reduce the dimensionality of inputs, speed up the training of the network and get better performance.