| At present, the combination of a variety of data mining technology so as to realize to the problem of fault diagnosis has become the development trend of the future research of fault diagnosis. The decision tree model with its advantages of strong readability, classification speed plays an irreplaceable role in the field of fault diagnosis. However, the noise of data in training data set and overfitting in the process of modeling seriously restrict the diagnosis efficiency of decision tree model. Rough set theory is related to on the premise of keep unchanged data classification ability, Improve tolerance of noise data models, thus expanding the scope of the application of data, Therefore, the combination of the two methods become a new research direction in the field of fault diagnosis.Based on rough set, the decision tree model was constructed and the diagnosis of failure data for the purpose, mainly do the following several aspects:First of all, the classical rough set theory and decision tree algorithm, are reviewed, analyzed these two methods combined with other means of data mining application.Secondly, after comparing three kinds of decision tree algorithm, using C4.5 algorithm to replace the previous proposed by rough set-ID3 algorithm of decision tree algorithm, so that the model can overcome the error due to different attribute values.Thirdly, the existing rough set decision tree model can't be the solution to noise data. we refer to a new concept of variable precision rough set theory, its application in the process of the choice of initial variables to the decision tree, so achieve the improved decision tree model can tolerate noise data to some extent.Finally, the improved rough set, the decision tree model is applied to the fault diagnosis data, and compared with the C4.5 decision tree model are compared, and the validation of the former than the latter, the decision tree smaller, better ability to predict, the generated rules more rich. |
PDF Full Text Request