蚁群与K均值聚类算法融合研究及其在用户分群中的应用

发布时间：2018-01-02 04:06

本文关键词：蚁群与K均值聚类算法融合研究及其在用户分群中的应用　出处：《西南科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：如今市场竞争日益激烈,企业面对数量巨大的消费者群体,传统的大众化营销的成本高且没有优势。对用户群体进行细分,针对不同用户群体进行精准的市场营销,能提高企业营销策略的效率,获得更好的营销效果。传统数据分析方法难以处理海量的用户数据,因此,采用数据挖掘技术中的聚类分析算法实现企业用户分群。论文给出了营销原理及数据挖掘技术相关理论说明,研究了聚类分析技术中K-means聚类算法和蚁群聚类算法的基本原理。分析了长虹社区论坛的用户数据,根据用户数据数量大、属性多的特点,提出适用于论坛用户聚类的改进算法Ant-K-means聚类算法。研究了K-means聚类算法对簇数量的选取和初始聚类中心的质量敏感以及蚁群聚类算法搜索时间长的问题解决方法。完成了蚁群聚类算法对数据集中抽取的部分样品数据的聚类,并将蚁群聚类获取的簇数量和簇中心作为输入参数,使用本文改进的基于信息素的K-means算法对全部数据进行快速聚类,获得最终聚类结果。将改进后的聚类算法应用到长虹智能电视的论坛用户分群中,根据用户的行为数据将用户聚类到不同的簇,对各簇的用户进行了分析与可视化。将改进算法应用到长虹论坛的用户分群中,实验结果显示分群效果良好,群体内用户特征相似性较高,验证了改进算法的正确性和有效性,为长虹智能电视的产品营销与用户服务提供了决策依据。
[Abstract]:Nowadays, the market competition is increasingly fierce, enterprises face a large number of consumer groups, traditional popular marketing has high cost and no advantages. Accurate marketing for different user groups can improve the efficiency of enterprise marketing strategies and achieve better marketing results. Traditional data analysis method is difficult to deal with massive user data, so. The clustering analysis algorithm of data mining technology is used to realize the clustering of enterprise users. The marketing principle and the related theory of data mining technology are given in this paper. This paper studies the basic principles of K-means clustering algorithm and ant colony clustering algorithm in clustering analysis technology, and analyzes the user data of Changhong Community Forum, according to the characteristics of large number of user data and many attributes. An improved Ant-K-means clustering algorithm for forum users is proposed. The sensitivity of K-means clustering algorithm to the selection of cluster numbers and the quality of initial clustering centers and ant colony clustering are studied. To solve the problem of long searching time, the ant colony clustering algorithm is used to cluster part of the sample data extracted from the data set. The number of clusters and cluster centers obtained by ant colony clustering are taken as input parameters, and the improved K-means algorithm based on pheromone is used to cluster all the data quickly. The improved clustering algorithm is applied to the forum user cluster of Changhong Intelligent TV, and the users are clustered into different clusters according to the user behavior data. The improved algorithm is applied to the user cluster of Changhong Forum. The experimental results show that the clustering effect is good and the similarity of user characteristics within the group is high. The correctness and effectiveness of the improved algorithm are verified, and the decision basis for the product marketing and customer service of Changhong Intelligent TV is provided.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP311.13

【参考文献】