CAJ | 학술논문

本文详细讨论了聚类问题若干算法的缺陷；并设计出一个新的有效算法：首先混合k-均值算法、最邻近算法、最大距离分裂算法构造出初始的启发式搜索算法；在此基础上运用改进的万有引力原理，按非线性的单减指数函数动态改变质量，由引力原理自动吸引质点聚类；计算双目标函数f1，f2，并给出寻优判据得到20个随机测试数据的聚类方案，最后全局优化验证算法不依赖于初始类中心的选择。
본문상세토론료취류문제약간산법적결함；병설계출일개신적유효산법：수선혼합k-균치산법、최린근산법、최대거리분렬산법구조출초시적계발식수색산법；재차기출상운용개진적만유인력원리，안비선성적단감지수함수동태개변질량，유인력원리자동흡인질점취류；계산쌍목표함수f1，f2，병급출심우판거득도20개수궤측시수거적취류방안，최후전국우화험증산법불의뢰우초시류중심적선택。
This paper discussed in detail about deficiencies of some clustering algorithm and con-structed a new effective method:First of all,mixed k-means algorithm,the nearest algorithm,the largest separatist distance algorithm and constructed the initial heuristic searching algorithm;Used im-provement gravitation principle and dynamic changed the particle quality according to the misalignment monotone decreasing exponential function,attracted the particle cluster automatically;Calculated the double objective function f1,f2,and gave the optimization criterion to obtain a good cluster plan for 20 random test data ;Finally confirmed the algorithm not to rely on the initial kind of center through the global optimization.