Feature Selection for Vibration Signal Based on Rough Set and MMAS

On the basis of dilation matrix, a new attribute reduction algorithm is put forward by applying the max-min ant system(MMAS) algorithm to finding reductions. Aiming at the problem of feature selection based on rough set theory, a comprehensive evaluation index is defined to evaluate the generalization capability and dimension of reductions. The reduction with the minimal index is regarded as the optimal feature subset, which can achieve the best compromise between generalization and dimension. By applying the algorithm to vibration signal, it is proved.