Aplicação de técnicas de mineração de dados em problemas de classificação de padrões

Abstract

The Knowledge Discovery in Databases (KDD) process, which includes the data mining phase, has been widely used as a tool to assist decision-making in areas such as banking credit and medical predictions. In this work this process is studied with the objective ofevaluating the use of data mining methods in areas of electrical engineering, considering data obtained from chromatography tests of power transformers. The data mining is applied for a classification of the types of transformers´s defects. The techniques that were studied in this work are neural networks and decision trees. The algorithms chosenin these techniques are, respectively, MLP´s network with resilient backpropagation algorithm for the training, simulated in Matlab, and the tree generated by the J4.8 algorithm, simulated in Weka. Chapter 2 presents a study about the KDD´s process, with the phases of the whole process and their activities. In Chapter 3 is given the data miningphases, highlighting its various aplications. In Chapters 4 and 5 the studies of the techniques and your algorithms are presented. Then are given two databases considered benchmark for the study validation and finally the algorithms are applied in the database of chromatography. In the following sections the results and conclusions are presented,where it is seen that the data mining can be applied to problems in the electrical engineering area, but must be made studies on the area of each database to be treated.