融合化学反应优化与K均值的文本数据聚类  

Text data clustering integrating chemical reaction optimization with K-means

作　　者：王琛[1] 董永权[2] WANG Chen;DONG Yong-quan(School of Information and Electronics Engineering,Jiangsu Vocational Institute of Architectural Technology,Xuzhou 221116,China;School of Computer Science and Tec nology,Jiangsu Normal University,Xuzhou 221116,China)

机构地区：[1]江苏建筑职业技术学院信电工程学院,江苏徐州221116 [2]江苏师范大学计算机科学与技术学院,江苏徐州221116

出　　处：《计算机工程与设计》2021年第8期2248-2256,共9页Computer Engineering and Design

基　　金：国家自然科学基金项目(61100167);江苏省自然科学基金项目(BK2011204)。

摘　　要：提出一种融合化学反应优化与K均值的文本聚类算法。结合K均值聚类的局部快速开发寻优能力和化学反应优化的全局勘探能力,以K均值聚类解集合作为化学反应优化的初始分子结构群,通过引入单分子碰撞、单分子分解、分子间碰撞和分子间合成4种化学反应操作,增加种群分子多样性;利用融合余弦相似度和欧氏距离的适应度函数评估分子优劣,在扩展搜索空间的基础上求解最优文本聚类。聚类测试结果表明,在多数数据集上,算法在适应度、准确率、精确率、召回率和F度量等指标上均表现更好。A text clustering algorithm fusing chemical reaction optimization with K-means clustering algorithm was proposed.Combined with the local rapid exploitation search optimization ability of K-means clustering and the global exploration ability of chemical reaction optimization,with the K-means clustering solution set as the initial group of molecular structure of chemical reaction optimization,by introducing four chemical reaction operations of single molecular collision,single molecules decompose,collision between moleculars and synthesis between moleculars,the molecular diversity of population was increased,moleculars were evaluated using the fitness function integrating cosine similarity and Euclidean distance,and the optimal text clustering was solved on the basis of expanding the search space.The results of clustering test show that,on most of datasets,the proposed algorithm performs better on some evaluation measures such as convergence,accuracy,precision,recall and F-measure.

关 键 词：文本聚类 化学反应算法 K均值聚类 矢量空间模型 适应度函数 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]