论文部分内容阅读
针对传统k-均值聚类算法事先必须获知类别数和难以确定初始聚类中心的缺点,建立了关于聚类中心和类别数k的双层规划模型,结合粒子群算法确定出聚类中心,通过在迭代过程中不断更新准则函数的方法搜索并确定出最佳类别数惫,基于所建模型,提出了一种改进的k-均值聚类算法,并将算法应用于冰脊表面形态分析中.结果表明,算法得到的聚类结果不但具有相邻类别边界清晰的优点,而且能够较好地反映出地理位置和生长环境对冰脊形成的影响.
For the traditional k-means clustering algorithm, the number of classes must be known in advance and the shortcomings of the initial clustering center are difficult to determine. A bilevel programming model about cluster center and number of classes k is established. The clustering center is determined by combining particle swarm optimization In the iterative process, the method of continuously updating the criterion function is used to search and determine the optimal number of clusters. Based on the model, an improved k-means clustering algorithm is proposed. The algorithm is applied to the ice ridge surface morphological analysis. The results show that the clustering results obtained by the algorithm not only have the advantages of clear boundaries of adjacent categories, but also reflect the influence of geographical location and growth environment on ice ridge formation.