ClassificaÃÃo supervisionada usando dados simbÃlicos de semÃntica modal

AUTOR(ES)

FÃbio CÃsar Donato Silva

DATA DE PUBLICAÇÃO

2007

RESUMO

The Symbolic Data Analysis (SDA) is a domain in the area of automatic discovery of knowledge that it aims at to develop methods for described data for variables that can assume as value lists of categories, intervals or distributions of probability. These variables allow to take in account the variability and/or uncertainty present in the data. This work presents a symbolic classifier of modal semantics for symbolic data of type interval. The considered classifier presents two basic stages, the learning and the allocation, where both need one step precedent of pre-processing that transforms the symbolic data of the type interval into symbolic data modal. Each example of the set of learning is described for a vector of intervals. After the pre-processing, each example starts to be described for a vector of distributions of weights. After the stage of learning, each group is also described for a vector of distributions of weights that summarize the information of the examples of the group. Each new example to be attributed to the one class (stage of allocation), represented for a vector of intervals, after the step of pre-processing starts to be described for a vector of distributions of weights. The allocation of an example to a class is carried through dissimilarity functions that compare pairs of vectors of distributions of weights. Some functions of dissimilarity of this type are considered in this work. The evaluation of the performance of this classifier is carried through the real application of the same the synthetic data sets in an experience Carlo Monte and reals data sets having used the technique of crossed validation leave-one-out. The performance is measured by the tax (average) of error of classification and by the time of execution of the stages of learning and classification. Moreover, the performance of this classifier was compared with the performance of a type classifier k nearest neighbors also to modal semantics. Through these examples, this work shows some of the interests of this classifier of modal semantics

ASSUNTO(S)

ciencia da computacao unsupervised classification dissimilarity functions symbolic data analysis classificaÃÃo supervisionada funÃÃes de dissimilaridade dados simbÃlicos modas analise de dados simbÃlicos modal symbolic data classificador simbÃlico modal modal symbolic classifier

Documentos Relacionados

• Modelos de regressÃo para dados simbÃlicos de natureza intervalar
• MÃtodo de Potencial para a classificaÃÃo supervisionada
• MÃtodos de agrupamento de dados simbÃlicos baseados em funÃÃes de proximidade
• Uma soluÃÃo em filtragem de informaÃÃo para sistemas de recomendaÃÃo baseada em anÃlise de dados simbÃlicos
• ClassificaÃÃo supervisionada de imagens SAR via redes neurais artificiais.