基于大数据挖掘的手机用户换机行为研究

发布时间：2018-01-06 01:09

本文关键词：基于大数据挖掘的手机用户换机行为研究　出处：《天津商业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着互联网技术的飞速发展和计算机应用的普及深化,手机行业的数据采集能力得到了前所未有的提高,获得并积累了浩如烟海的移动通讯大数据。通过大数据挖掘技术,研究手机用户的换机特征及换机行为的影响因素具有十分重要的意义,它能够使我们深入地把握用户需求,实现精准营销。基于这一背景,本文开展了基于大数据挖掘的手机用户换机行为研究。首先,依托某省公司电信运营商经分系统,获得了某段时期内的手机用户换机数据,在界定手机用户换机行为内涵、构建指标体系的基础上,按照简单随机抽样的方法,得到了Kullback-Leibe信息量达到0.99995以上的样本,并对样本数据进行了数据管理、缺失值处理、离群点检测等工作,保证了数据质量;然后,利用可视化技术对换机前后的手机品牌特征、手机性能特征及用户特征进行了探索性分析;接着,本文分别采用决策树、Adaboost、Bagging及随机森林算法对数据进行训练,建立预测模型,得出了各指标对换机行为的影响程度,并根据十折交叉验证的平均误判率对不同的模型结果进行了分析和比较,其中Adaboost及随机森林模型较优;最后,综合各模型得出的指标重要性排序,结合实际,总结了各指标对换机行为的影响情况,并提出了手机厂商要注重品牌建设、加大技术投入和创新力度等建议。本文以大量的真实数据为第一手资料,采用定性分析与定量分析相结合的研究方法,通过可视化技术对手机用户的换机行为进行了深入的研究,在此基础上,引入了数据挖掘中的多种算法分析了换机行为的影响因素,并提出了一些生产、销售建议,具有一定的创新性。但是本文在指标选取、数据抽样过程中还存在一些有待改进的地方,可在以后的研究工作中加以完善。
[Abstract]:With the rapid development and popularization of computer Internet technology applications deepening, data acquisition ability of the mobile phone industry has been hitherto unknown rise, and accumulated a large data mobile communication. Through the multitude of data mining technology, has very important significance to study the factors affecting the behavior of replacement of mobile phone users characteristics and replacement, it can make we deeply grasp the needs of users, to achieve precision marketing. Based on this background, this paper carried out the research of data mining based on the behavior of mobile phone users switch. First of all, relying on a company of telecom operators to manage the system, obtained within a certain period of replacement Mobile phone users, mobile phone users switch behavior in defining the connotation, construction of basic indicators the system, in accordance with the simple random sampling method, the information content of Kullback-Leibe reached more than 0.99995 of the sample, and the number of samples According to the data management, missing data processing, outlier detection work, to ensure the quality of data; then, mobile phone brand characteristics of replacement before and after the application of visualization technology, an exploratory analysis of mobile phone performance characteristics and user characteristics; then, this paper uses decision tree, Adaboost, Bagging and random forest algorithm for training the data to establish the prediction model, the influence degree of each index on the replacement behavior, and according to the different model results are analyzed and compared with the average error rate of ten fold cross validation, including Adaboost and random forest model is better; finally, the index of importance, the comprehensive model combined with practice, summed up the analysis of the influence of replacement behavior, and put forward the mobile phone manufacturers to focus on brand building, increase technology investment and innovation and so on. Based on a large number of real data for the One hand, adopting the research method of combining qualitative analysis and quantitative analysis, through the switching of visualization technology to the mobile phone users are studied, on this basis, the introduction of a variety of data mining algorithms to analyze the factors affecting the replacement behavior, and puts forward some suggestions of production, sales, innovative sure. But based on the index selection, there are still some room for improvement in the process of sampling data, which could be improved in future studies.

【学位授予单位】：天津商业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;C913.4

【参考文献】