A Fault Diagnosis Method Combining Rough Sets And Neural Network

Rough sets and neural networks are two common techniques applied to data mining problems in order to inprove diagnosis precision and decreasing misinformation diagnosis. Integrating the advantages of two approaches, this paper presents a hybrid system to extract efficiently classification rules from decision table. The target is mainly to remove redundant information and seek for reduced decision tables which to obtain he minimum fault feature subset. The neural networks adopted were of the feed-forward variety with one hidden layer. They were trained using back-propagation. The effectiveness of our approach was verified by the experiments comparing with traditional rough set and neural network approaches, and can detect the composed faults while keep good robustness.