摘要
为探寻区域性小麦品质聚类的适宜算法,针对经典K-Means(KM)算法对初始聚类中心的敏感问题,以我国主要冬麦区为研究实例,探讨了两种初始中心点改进算法对大规模小麦品质数据集的适应性,综合距离与密度两因素,提出了一种基于密度参数和邻域半径的优化初始中心点小麦品质聚类算法。相对KM算法及文献改进算法,所提算法可较为准确地提取数据集高密度区域的初始中心点,聚类过程及性能对静态簇与非静态簇两种不同迭代方案相对不敏感。实验结果验证了算法的有效性和可行性,在收敛性能及稳定性方面具有一定的优势。
To explore the suitable algorithm for clustering wheat quality regionally,it discusses the adaptability of two optimized initial center points algorithms based on the case study of main winter wheat region in China aiming at the problem that traditional K-means is sensitive to the initial centers.A method is proposed to optimize the method of choice initial center points through computing the density of objects and the distance between objects which is adjusted by the adjustment coefficient.Compared with original algorithm and optimized algorithms,it has the advantages of effectively capturing the initial center points scattered over the high-density region of the large-scale data set,and representing relatively insensitivity to the clustering schemes statically or astatically.The experimental results show effectiveness and feasibility for clustering wheat quality regionally using the proposed algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2011年第36期34-37,72,共5页
Computer Engineering and Applications
基金
国家自然科学基金(the National Natural Science Foundation of China under Grant No.30971697)
中央高校基本科研业务费(No.KYZ201005)
南京农业大学人文社科基金(No.sk09016)
关键词
区域聚类
小麦品质
优化初始中心
适宜性
region clustering
wheat quality
optimized initial center
suitability
