Improved Feature Selection Algorithm Based on Maximal Nearest-Neighbor Rough Approximation

The feature selection algorithm based on maximal nearest neighbor rough approximation can not only deal with the mixed data, but also avoid the choice of the parameter values in the feature selection algorithm based on neighborhood rough sets. And it reduces the judgement of the sample. But the evaluation standard of this method only considers the importance of a single attribute which is relative to the result of the decision while calculating the importance of the attribute. It ignores the influence of the interaction between the attributes on the result of decision. So this paper sets up the new evaluation standard which is considered the influence of the attributes, and a forward greedy feature selection algorithm is constructed. Experiments show that the proposed algorithm can not only select fewer features, but also improve the accuracy of the classification.