若干密度聚类算法改进及电商网站评价应用

发布时间：2018-04-11 22:45

本文选题：聚类分析 + DBSCAN　；参考：《吉林财经大学》2016年硕士论文

【摘要】：聚类分析是数据挖掘的重要技术,是获取知识的重要方法和手段。DBSCAN算法作为一种基于密度的聚类分析方法,可以在包含噪音点的数据中发现任意形状的聚类且不受噪音点的影响。但该算法也存在以下缺点:(1)输入参数需要先验知识人工设置;(2)对输入参数敏感,特别是处理密度不均的数据时,得到的聚类结果不佳。针对DBSCAN算法的参数敏感性问题,本文提出基于数据划分的DBSCAN算法。该聚类算法的基本思想:首先采用人工鱼群算法和k-means划分相结合的方法实现数据集的划分;然后在数据划分的基础上,引入变异系数优化参数Eps取值,使用DBSCAN算法分别聚类;最后得到最终的聚类结果。为了验证算法的有效性和有用性,本文将改进的DBSCAN算法应用于企业电子商务网站的评价。随着互联网和电子商务的迅速普及和发展,大多数企业在互联网上建立自己的电子商务网站,提升企业竞争力。然而,由于没有统一的衡量评价标准,企业对电子商务网站缺乏正确的认识,需要对企业电子商务网站进行合理分析和评价,帮助企业提高电子商务网站建设水平。另外,电子商务网站的数据具有密度不均的分布特性,因此,适合用改进的DBSCAN聚类算法对企业电子商务网站进行聚类分析和评价。企业电子商务网站评价问题的成功解决,也将有助于DBSCAN算法在实际应用方面进行更加深入的研究。
[Abstract]:Clustering analysis is an important technology of data mining. It is an important method and means of acquiring knowledge. DBSCAN algorithm is a density-based clustering analysis method.Clusters of arbitrary shapes can be found in data containing noise points and are not affected by noise points.However, the algorithm also has the following disadvantages: 1) the input parameters need to be manually set by a prior knowledge) and the algorithm is sensitive to the input parameters, especially when dealing with data with uneven density, the clustering results obtained are not good.Aiming at the parameter sensitivity of DBSCAN algorithm, this paper proposes a DBSCAN algorithm based on data partitioning.The basic idea of this clustering algorithm is: firstly, the method of combining artificial fish swarm algorithm and k-means partition is used to realize the partition of data sets, and then, on the basis of the data partition, the variation coefficient is introduced to optimize the parameter Eps, and the DBSCAN algorithm is used to cluster the data separately.Finally, the final clustering results are obtained.In order to verify the validity and usefulness of the algorithm, this paper applies the improved DBSCAN algorithm to the evaluation of enterprise e-commerce websites.With the rapid popularization and development of the Internet and e-commerce, most enterprises set up their own e-commerce websites on the Internet to enhance the competitiveness of enterprises.However, due to the lack of a unified evaluation standard and lack of correct understanding of e-commerce websites, it is necessary to analyze and evaluate them reasonably to help them improve the level of construction of e-commerce websites.In addition, the data of E-commerce website is distributed unevenly, so it is suitable to use the improved DBSCAN clustering algorithm to analyze and evaluate the E-commerce website.The successful solution of the evaluation problem of enterprise e-commerce website will also be helpful for the further study of DBSCAN algorithm in practical application.
【学位授予单位】：吉林财经大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.13;F724.6

【相似文献】