动态聚类算法从本质上讲是单目标组合优化算法,一般需要事先给定目标分类数和初始聚类中心,且初始聚类中心的选择对数据划分结果影响较大。为了解决该问题,提出将产状数据的划分问题转化为多目标优化问题,并采用小生境Pareto遗传算法进行求解。针对聚类问题的特殊性,采用基于链表的编码方案,并建议相应的遗传操作算子;通过引入小生境技术和Pareto支配集理论,仅通过一次求解可由Pareto支配集给出对应于不同目标组数的最优分组结果,而且不用事先给定目标组数以及初始聚类中心。最后,将算法应用于三峡船闸高边坡岩体实测不连续面产状数据的划分,得到较为符合实际的优势结构面分组。
Clustering method is a single-objective optimal method in nature, a prior specified number of clusters and the initial cluster centroids must be given in advance, and different choices of initial guesses of cluster centroids can lead to different partitions of the same data. The new representation proposed in this paper deals with the partitional clustering problem by regarding it as a multi-objective optimal problem; in this approach the niche Pareto genetic algorithm is used to solve the problem. Aiming at clustering problem, a linked-list based encoding scheme and accordingly genetic operators are presented. With the introduction of niche technique and Pareto dominant set theory, the optimal partitions for all possible numbers of clusters in the Pareto optimal set returned by a single GA run are obtained. The performance of the proposed approach has been tested using artificial data and the data form real rock mass of the shiplock high slope of the Three Gorges Project. The obtained results are promising and demonstrate the applicability and effectiveness of the proposed approach.