CAJ | 학술논문

传统的k-means聚类算法常陷入局部最优，需要事先输入聚类数，这样会造成原有算法失效或聚类结果不准确。在研究现有聚类算法的基础上，使用ε-最近邻法剔除孤立点，提出一种改进的基于模拟退火算法的、具有自适应功能的k-means聚类算法。实验结果证明，提出的算法是可行的、有效的。
전통적k-means취류산법상함입국부최우，수요사선수입취류수，저양회조성원유산법실효혹취류결과불준학。재연구현유취류산법적기출상，사용ε-최근린법척제고립점，제출일충개진적기우모의퇴화산법적、구유자괄응공능적k-means취류산법。실험결과증명，제출적산법시가행적、유효적。
The problem of traditional k-means clustering algorithm is that it often falls into local optimum, and needs to input clustering parameters ahead of time, which causes some disabled or inaccurate clustering results. On the basis of current clustering algorithm research, this paper proposes a kind of improved k-means clustering algorithra based on simulated annealing. The algorithm is adaptive and may remove the oufliers by ε-nearest neighbor method. And its feasibility and effectiveness are proved by the experiment results.