基于改进磷虾群算法的多目标文本聚类方法  

Multi-objective text clustering method based on improved krill herd algorithm

作　　者：菊花 JU Hua(College of Education,Inner Mongolia Normal University,Hohhot 010020,China)

机构地区：[1]内蒙古师范大学教育学院,内蒙古呼和浩特010020

出　　处：《计算机工程与设计》2022年第6期1694-1703,共10页Computer Engineering and Design

基　　金：教育部人文社科研究规划基金项目(20XJA740002)。

摘　　要：提出融合K均值与改进磷虾群算法的多目标文本聚类算法。利用K均值的局部快速寻优和改进磷虾群的全局搜索能力,以K均值聚类解作为改进磷虾群的初始种群,引入遗传交叉和变异改善个体多样性,提升全局搜索能力;通过磷虾种群的诱导运动、觅食运动和随机扩散进行位置更新,引入余弦相似度和欧氏距离的多目标适应度函数评估磷虾位置优劣,搜索全局最优解。通过基准数据集实验确定磷虾群算法的关键参数,进行系统聚类测试,实验结果表明,该算法在聚类指标上表现更佳,聚类准确性更高,收敛速度更快。A multi-objective text clustering algorithm integrating improved krill herd algorithm with K-means clustering was pre-sented.The rapid local exploitation of K-means clustering was combined with the global exploration ability of improved krill herd algorithm.With the solution set generated by the K-means clustering as the initial population of improved krill herd algorithm,the genetic crossover and mutation was introduced to enhance the diversity of individuals and the global exploration ability was promoted.The position was updated by the induce movement,foraging movement and random diffusion in the krill herd.A multi-objective fitness function integrating cosine similarity and Euclidean distance was introduced to evaluate the quality of krill’s position and to search for the global optimum.The key parameters of improved krill herd algorithm were determined by the experiments of benchmark datasets.On this basis,a systematic clustering test was conducted.The results show that the proposed algorithm performs better on the clustering index,and has higher clustering accuracy and convergence speed.

关 键 词：文本聚类 K均值算法 磷虾群算法 遗传算子 多目标聚类 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]