Study On Model Of Data Mining Of Vehicle Faults Diagnosis

In this paper the basic theory of data mining is briefly introduced. For particularity and complexity of vehicle faults diagnosis, it is necessary to integrate rough set theory with neural network. So a framework of vehicle faults diagnosis based on RS and NN has been proposed and illustrated in an application for the faults diagnosis of gearbox. Continuous attributes are discretized by means of Fuzzy c-means, Kohonen neural network and K-means. Corresponding rules sets are compared. The result shows discretized ways have an important effect on rules, so right discretized ways should be considered in use. Moreover, reduces of RS not only simplify NN structure, but also increase NN efficiency. Finally the paper considers integrating of RS and NN is promising.