An Attribute Reduction Based on Genetic Algorithm and Rough Sets

To solve the problem that low efficiency and slow convergence speed of traditional attribute reduction algorithm, we propose an attribute reduction algorithm which based on the genetic algorithm and rough sets. To obtain the minimum attribute reduction, attribute dependence and hamming distance as constrains is introduced in population initialization. When the fitness function is designed, the average attribute importance is introduced as the correction factor, and the fitness function is dynamically adjusted. The improved adaptive crossover and mutation probability are adopted, and in the cross-operation, a small-scale competition strategy is used. Experimental results prove the efficiency of the proposed algorithm in attribute reduction for high dimensionality and big data.