Catálogo de publicaciones - libros

We propose an approach to subgroup discovery using distribution rules (a kind of association rules with a probability distribution on the consequent) for numerical properties of interest. The objective interest of the subgroups is measured through statistical goodness of fit tests. Their subjective interest can be assessed by the data analyst through a visual interactive subgroup browsing procedure.

Ontologies provide a common layer which plays a major role in supporting information exchange and sharing. In this paper, we focus on the ontological concept extraction process from HTML documents. We propose an unsupervised hierarchical clustering algorithm namely “Contextual Ontological Concept Extraction” (COCE) which is an incremental use of a partitioning algorithm and is guided by a structural context. This context exploits the html structure and the location of words to select the semantically closer cooccurrents for each word and to improve the words weighting. Guided by this context definition, we perform an incremental clustering that refines the words’ context of each cluster to obtain semantic extracted concepts. The COCE algorithm offers the choice between either an automatic execution or an interactive one. We experiment the COCE algorithm on French documents related to the tourism. Our results show how the execution of our context-based algorithm improves the relevance of the clusters’ conceptual quality.

This paper presents a full scaled application of induction trees for non-classificatory purposes. The grown trees are used for highlighting regional differences in the women’s labor participation, by using data from the Swiss Population Census. Hence, the focus is on their descriptive rather than predictive power. Trees grown by language regions exhibit fundamental cultural differences supporting the hypothesis of cultural models in female participation. The explanatory power of the induced trees is measured with deviance based fit measures.

Microarray technology produces vast amounts of data by measuring simultaneously the expression levels of thousands of genes under hundreds of biological conditions. Nowadays, one of the principal challenges in bioinformatics is the interpretation of huge data using different sources of information.

We propose a novel data analysis method named CGGA (Co-expressed Gene Groups Analysis) that automatically finds groups of genes that are functionally enriched, i.e. have the same functional annotations, and are co-expressed.

CGGA automatically integrates the information of microarrays, i.e. gene expression profiles, with the functional annotations of the genes obtained by the genome-wide information sources such as Gene Ontology (GO).

By applying CGGA to well-known microarray experiments, we have identified the principal functionally enriched and co-expressed gene groups, and we have shown that this approach enhances and accelerates the interpretation of DNA microarray experiments.

In this paper, we propose a method for discovering hidden information from large-scale item set data based on the symmetry of items. Symmetry is a fundamental concept in the theory of Boolean functions, and there have been developed fast symmetry checking methods based on BDDs (Binary Decision Diagrams). Here we discuss the property of symmetric items in data mining problems, and describe an efficient algorithm based on ZBDDs (Zero-suppressed BDDs). The experimental results show that our ZBDD-based symmetry checking method is efficiently applicable to the practical size of benchmark databases.

Category-based induction is a kind of inductive reasoning in which the premise and the conclusion of the argument is in the form Rips and Osherson et al. investigated the argument strength of category-based induction, and Lopez et al. showed that there are differences of the acceptability of category-based induction between infants and growing-ups. There are two problems in their analysis. One is the ambiguity of the difference between categories and individuals, and the other is the reason of the changes of the acceptability in developmental process of logical inference. In this paper we give mathematical models category-based induction and, based on the models, propose a hypothesis which explains the reason of the problems.

Constructing evaluation functions with high accuracy is one of the critical factors in computer game players. This construction is usually done by hand, and deep knowledge of the game and much time to tune them are needed for the construction. To avoid these difficulties, automatic construction of the functions is useful. In this paper, we propose a new method to generate features for evaluation functions automatically based on game records. Evaluation features are built on simple features based on their frequency and mutual information. As an evaluation, we constructed evaluation functions for mate problems in shogi. The evaluation function automatically generated with several thousand evaluation features showed the accuracy of 74% in classifying positions into mate and non-mate.

In this paper, a new algorithm (RESOP) is introduced. This algorithm reduces databases in two directions: editing examples and feature selection simultaneously. Ordered projections techniques have been used to design RESOP taking advantage of symmetrical ideas for two different task. Experimental results have been made with UCI Repository databases and the performance for the latter application of classification techniques has been satisfactory.

The solution of multi-agent system could be applied for air pollution monitoring and control systems modelling in the context of extending the area of web based applications to environmental systems. As the intelligent agents that compose such a multiagent system need to communicate between them and also with external agents they must share parts of their ontologies or they must identify the correspondent common terms. In this paper, we focus on the topic of ontology mapping in such a multi-agent system.

This work shows, using bivariate continuous artificial domains, the relation that seems to exist between some measures based on the information theory and the expected classification error.

The relations that seem to be found in this work could be applied to the improvement of the classifiers which assign probabilities to each class value. They also could be used in other tasks related to the supervised classification such as feature subset selection or discretization.