Abstract: KNN is one of the arithmetic which based on the instance.The number of training samples influence on the classification performance of KNN.Reasonable sample cut can improve the efficiency of classification.This paper proposes an improved density-based KNN model under the clustering conditions.Firstly, basing on the types, we used the clustering method to choice the samples in order to reduce the noise samples.Then we weighted samples basing on sort density and overcame the defect that the big class, high density of training samples dominated in the KNN.The result of test shows that the improved KNN classification algorithm improves the efficiency of KNN classification.