自适应截断密度峰值聚类欠采样研究

he unbalance of data distribution aggravates the difficulty of classification. In order to reduce the influence caused by the unbalanced distribution of classes, this paper proposes an adaptive peak density clustering under-sampling algorithm by combining the advanced peak density clustering and resampling methods, the problem that the truncation distance of density peak clustering depends on artificial setting is solved. In this algorithm, the optimal truncation distance is automatically selected by random forest and Bayesian optimization algorithm, which optimizes the classification effect after undersampling. At the same time, by means of Bayesian optimization, it can reduce the running time complexity and be more efficient. In order to verify the effectiveness of the improved algorithm, this paper compares the sampling effect of the new algorithm with the original ones, so as to verify the effectiveness of the improved algorithm from the practical level.