MÃtodos de visualizaÃÃo de informaÃÃes na descoberta de conhecimento em bases de dados

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

The process of Knowledge Discovery in Databases (KDD) aims to support the process of decision making through the automatic extraction of strategic, useful and hidden knowledge embedded in large databases. The discovered knowledge needs to be analyzed and easily understood by users and managers in general, in order to make it relevant to solve problems in daily operations or to planning actions in the context of the problem being studied. The discovered knowledge may be represented in many different ways. However, some of these forms often are not clearly understood by users or do not allow detailed analysis or validation of new hypothesis. To help the interpretation of the results obtained by the mining process, graphical Information Visualization techniques have significantly contributed: for an intelligent representation of large data sets, for the use of statistical techniques in data analysis and for the visual manipulation of data. The application of these techniques over the KDD process is called Visual Data Mining. The main goals of this work are the investigation of visual information techniques when applied to the KDD process; the modeling and development of a software tool, the VisualDATAMINER, which implements visual data mining methods and techniques adapted to the interpretation of the mined results; and the accomplishment of a case study in a large scale real world problem in the credit risk assessment domain, for the validation of the developed software. The VisualDATAMINER makes possible the interpretation of induction rules, allows integration with other data mining software, facilitates the visualization of the mined results in many different ways and allows user manipulation of these views through interaction methods. Developed in Java, the VisualDATAMINER tool presents all the benefits of an object oriented software such as reusability, maintainability, and encapsulation. An experimental investigation was performed using a large scale credit risk assessment database, and the results observed demonstrated the refinement of the knowledge extracted through the application of the visualization techniques and interaction methods proposed in the software, assuring the efficiency and benefits produced by VisualDATAMINER tool

ASSUNTO(S)

visual data mining visualization of information visualizaÃÃo de informaÃÃes mining of data visual mining date mineraÃÃo de dados ciencia da computacao

ACESSO AO ARTIGO

Documentos Relacionados