Catálogo de publicaciones - libros

In many classification problems, some classes are more important than others from the users’ perspective. In this paper, we introduce a novel approach, , to address this issue by modeling class importance through weights in the [0,1] interval. We also propose novel metrics to evaluate the performance of classifiers in a weighted classification context. In addition, we make some modifications to the ART classification model [1] in order to deal with weighted classification.

We discuss generalizations of the basic notion of a template defined over information systems using indiscernibility relation. Generalizations refer to the practical need of operating with more compound descriptors, over both symbolic and numeric attributes, as well as to a more entire extension from equivalence to tolerance relations between objects. We briefly show that the heuristic algorithms known from literature to search for templates in their classical indiscernibility-based form, can be easily adapted to the case of tolerance relations.

Theory of Rough Sets provides good foundations for the attribute reduction processes in data mining. For numeric attributes, it is enriched with appropriately designed discretization methods. However, not much has been done for symbolic attributes with large numbers of values. The paper presents a framework for the symbolic value partition problem, which is more general than the attribute reduction, and more complicated than the discretization problems. We demonstrate that such problem can be converted into a series of the attribute reduction phases. We propose an algorithm searching for a (sub)optimal attribute reduct coupled with attribute value domains partitions. Experimental results show that the algorithm can help in computing smaller rule sets with better coverage, comparing to the standard attribute reduction approaches.

After the tragic events of 9/11, the concern about national security has increased significantly. However, law enforcement agencies, particularly in view of current emphasis on terrorism, increasingly face the challenge of information overload and lack of advanced, automated techniques for the effective analysis of criminal and terrorism activities. Data mining applied in the context of law enforcement and intelligence analysis, called Investigative Data Mining (IDM), holds the promise of alleviating such problems. An important problem targeted by IDM is the identification of terror/crime networks, based on available intelligence and other information. In this paper, we present an understanding to show how IDM works and the importance of this approach in the context of terrorist network investigations and give particular emphasis on how to destabilize them by knowing the information about leaders and subgroups through hierarchical structure.

The development of applications that use the different data sources available in organizations require to solve a data integration problem. Most of the methodologies and tools that simplify the task of finding an integrated schema propose conventional object-oriented solutions as the basis for building a global view of the system. As we will see in this work, the use of the conventional object-oriented data model is not as appropriate as we would like when dealing with data variability and we present a novel typing framework, lazy typing, that can be used for obtaining a global schema in the data integration process. This typing framework eases the transparent development of applications that use this integrated schema and reconcile data.

This paper provides the methodologies to extract the road horizontal alignment design elements using the acquisition data from the Global Positioning System (GPS) and Inertial Navigation System (INS). For this study, highly accurate GPS/INS data from the RoSSAV (Road Safety Survey and Analysis Vehicle) were collected, and also extraction algorithm of road horizontal alignment design elements was proposed according to the statistical inference.

The amount and quality of feedback provided to the learner has an impact on the learning process. Personalized feedback is particularly important to the effective delivery of e-learning courses. E-learning delivery methods such as web-based instruction are required to overcome the barriers to traditional-type classroom feedback. Thereby, the feedback for a learner should consist not only of adaptive information about his errors and performance, but also of adaptive hints for the improvement of his solution. Furthermore, the tutoring component is required to individually motivate the learners. In this paper, an adaptive assessment and feedback process model for personalized e-learning is proposed and developed for the purpose of maximizing the effects of learning.

Most option pricing methods use mathematical distributions to approximate underlying asset behavior. However, it is difficult to approximate the real distribution using pure mathematical distribution approaches. This study first introduces an innovative computational method of pricing European options based on the real distributions of the underlying asset. This computational approach can also be applied to expected value related applications that require real distributions rather than mathematical distributions. The contributions of this study include the following: a) it solves the risk neutral issue related to price options with real distributions, b) it proposes a simple method adjusting the standard deviation according to the practical need to apply short term volatility to real world applications and c) it demonstrates that modern databases are capable of handling large amounts of sample data to provide efficient execution speeds.

This research is exploring the extent to which providing traffic information on VMS affects drivers’ route choice behaviour. The information include extra delay and charges. Three different charging regimes were tested. Stated preference(SP) surveys were conducted and route choice logit models were estimated. The results show that drivers’ route choice is affected by length of delay and by road user charges on VMS. The fixed charges may be most likely to induce drivers to change their behaviour. Drivers value delay time more highly and they become increasingly sensitive to delay time as it increases.

Daegu Metropolitan City is currently in the process of implementing an Intelligent Transportation Systems (ITS) basic plan in order to establish these systems and the foundation of basic services, in addition to setting establishment goals based on the national basic plan of ITS. Some criteria have proven to be very effective at determining the priorities of ITS services, measuring their contribution to solving transportation problems, identifying the services preferred by users, and evaluating ITS systems and related technologies. In this study, the authors prioritize six ITS services using the Analytic Network Process (ANP), which considers mutual dependence between the evaluation items and alternatives. The Analytic Hierarchy Process (AHP), meanwhile, is a one-way process that does not consider the independence of feedback from the services. According to the results of the super decisions ratings, the Regional Traffic Information Center System was chosen to be the top priority project followed by the Urban Arterial Incident Management System and the Bus Information System.