人工鱼群K-means算法及其在通信客户细分中的应用研究

发布时间：2018-05-27 12:41

本文选题：聚类分析 + K-means算法　；参考：《合肥工业大学》2017年硕士论文

【摘要】：K-means算法对大型数据集的处理有较好的扩展性,较高的效率。然则Kmeans算法存在不足之处,它的不足之处就是聚类的效果很大程度上受初始聚类中心选择的影响,而且整体的开销容易受到数据初始量大小的影响。所以为了弥补传统聚类算法在聚类分析中的一些缺点,一些研究人员借鉴仿生学,从生物进化的机理中将一些智能算法与聚类算法融合,优化聚类过程,提升算法效能。而人工鱼群算法(Artificial Fish Swarm Algorithm,AFSA)作为一种仿生优化算法,其本身拥有对初始参数要求低、结构简单和寻优速度快等优点,可用于许多优化模型的求解。因此将人工鱼群算法用来优化K-means算法将能够有效提升聚类效果。论文在深入分析K-means算法和人工鱼群算法的国内外研究现状基础上,在深入研究K-means算法和人工鱼群算法的相关理论和方法后,针对K-means算法不足之处,提出用人工鱼群算法来优化K-means算法的聚类过程,通过构造人工鱼个体、确定算法目标函数,提出用于解决聚类问题的人工鱼群K-means算法。论文设计相关对比实验,通过编程将人工鱼群K-means算法实现,并与K-means算法在相同条件下处理相关数据集,运用最优值、最优值次数、准确度等相关评估准则,对比评估结果并对算法性能进行验证。同时将此算法运用到通信领域的客户细分中,根据实验分组结果制定相应营销策略。
[Abstract]:K-means algorithm has good expansibility and high efficiency for processing large data sets. However, the Kmeans algorithm has some shortcomings, which is that the effect of clustering is greatly affected by the selection of the initial clustering center, and the overall cost is easily affected by the size of the initial data. So in order to make up for some shortcomings of traditional clustering algorithm, some researchers learn from bionics, combine some intelligent algorithms with clustering algorithm from the mechanism of biological evolution, optimize the clustering process and improve the efficiency of the algorithm. As a bionic optimization algorithm, artificial Fish Swarm algorithm (AFSA) has the advantages of low requirements for initial parameters, simple structure and fast optimization speed, which can be used to solve many optimization models. Therefore, using artificial fish swarm algorithm to optimize K-means algorithm can effectively improve the clustering effect. On the basis of deeply analyzing the research status of K-means algorithm and artificial fish swarm algorithm at home and abroad, after deeply studying the relevant theories and methods of K-means algorithm and artificial fish swarm algorithm, this paper aims at the deficiency of K-means algorithm. An artificial fish swarm algorithm is proposed to optimize the clustering process of K-means algorithm. By constructing artificial fish individuals and determining the objective function of the algorithm, an artificial fish swarm K-means algorithm is proposed to solve the clustering problem. This paper designs the correlation contrast experiment, realizes the artificial fish swarm K-means algorithm through the programming, and processes the correlation data set with the K-means algorithm under the same condition, uses the best value, the best value times, the accuracy and so on correlation appraisal criterion. The evaluation results are compared and the performance of the algorithm is verified. At the same time, the algorithm is applied to the customer segmentation in the field of communication, and the corresponding marketing strategy is formulated according to the experimental results.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F274;F626

【参考文献】