摘要
粗糙C均值聚类算法采用随机选取质心的方法,会导致聚类算法过早陷入局部最优;簇心更新中采用固定权值降低了聚类精度。针对此类问题,结合果蝇和粗糙C均值聚类两种算法,提出一种改进算法。该算法从三个方面进行了改进:一是为克服传统果蝇算法中固定飞行半径带来的影响,给出一种自适应寻优步长策略的果蝇优化算法,提高果蝇优化算法的搜索精度;二是构造对应味道浓度值的目标函数,利用目标函数值引导果蝇进行位置更新,把最优味道浓度值的果蝇位置作为新的聚类中心进行次迭代;三是设计了一种动态调整簇心更新中上下近似权重和阈值的方案。最后通过UCI标准数据集对算法进行比对分析,实验结果证明了改进后算法的可行性和有效性。
Rough C-means clustering algorithm uses the method of random centroid selection,which will cause the clustering algorithm to fall into local optimal prematurely.The clustering accuracy is reduced by using fixed weight in cluster center updating.To solve this problem,an improved method is proposed by combining drosophila and rough C-means clustering.The new algorithm is improved from the following three aspects.Firstly,to overcome the influence of fixed flight radius in traditional drosophila algorithm,a drosophila optimization algorithm with adaptive uber length search strategy is proposed to improve the search accuracy of drosophila optimization algorithm.Secondly,we construct the objective function corresponding to the flavor concentration value,and use the objective function value to guide the drosophila to update the location.The location of the optimal flavor concentration value is used as the new clustering center for the next iteration.Thirdly,a scheme dynamically adjusting the approximate weights and thresholds in cluster core updating is designed.Finally,the algorithm is compared and analyzed by UCI standard data set,and the experimental results prove the feasibility and effectiveness of the improved algorithm.
作者
周浩岩
叶军
谢立
卢岚
李兆彬
ZHOU Haoyan;YE Jun;XIE Li;LU Lan;LI Zhaobin(School of Information Engineering,Nanchang Institute of Technology,Nanchang 330022,China;Jiangxi Province Key Laboratory of Water Information Cooperative Sensing and Intelligent Processing,Nanchang Institute of Technology,Nanchang 330022,China)
出处
《南昌工程学院学报》
CAS
2023年第4期79-86,共8页
Journal of Nanchang Institute of Technology
基金
江西省教育厅科学技术研究项目(GJJ211920)
国家自然科学基金资助项目(61562061)。
关键词
粗糙C均值
果蝇算法
适应度函数
权重因子
飞行策略
rough C-means
drosophila algorithm
fitness function
weight factor
flight strategy
