电信业客户细分研究

发布时间：2018-06-15 03:51

本文选题：电信业 + 特征分析　；参考：《浙江工商大学》2017年硕士论文

【摘要】：当前,电信业的竞争变得越来越激烈,而且随着智能手机和互联网应用的普及,一些网络聊天工具像微信、QQ这类通讯软件对电信业的传统业务如短信业务和语音业务带来了一定的影响,因此,进行客户细分,识别有价值的潜在客户变得尤为重要。本文拓宽了以单一指标如客户价值来进行客户细分的指标体系,结合互联网的深入应用给企业和客户带来的变化,最终确立了一套贴合实际的指标体系。在确立客户细分指标体系时,本文将数据挖掘的思想融入其中。首先基于可获得的大量原始数据,对其进行数据预处理,再结合电信业的行业特征确立了上网客户和非上网客户的细分指标体系。对于非上网客户的指标体系主要包括通话、消费等客户的传统特征,而对于上网客户,则主要将客户的上网行为如流量的使用、对APP的浏览行为纳入指标体系。此外,本文从多个角度比较全面的对电信业的客户特征进行了分析。在分析过程中,将上网客户和非上网客户分开,分别进行特征分析。不仅包括传统的统计特征,如年龄、性别、套餐、终端、通话行为等,还创新性的从多个方面,对上网客户的APP使用行为进行了分析,并将结果进行了可视化的展示。在模型改进方面,为了克服Kohonen SOM算法和K-Means算法的缺点,本文将KohonenSOM算法和K-Means算法进行了结合,创建了 Kohonen SOM+K-Means聚类分析模型。Kohonen SOM首先进行一次初始聚类,确定K值的个数,将其结果作为K-Means聚类的初始输入,最终将上网客户细分成了"普通人"、"社交王"、"阅读迷"、"生活控'"和"购物狂" 5类,将非上网客户细分成了 6类,分别是:"不活跃客户群"、"长途夜间活跃客户群"、"低端主动客户群"、"高语音亲情网内客户群"、"较高消费本地客户群"和"高消费漫游客户群'"。根据不同客户群的客户特征,提出了针对该电信运营公司精准营销的策略及建议。本文还基于聚类结果筛选出的高价值客户对客户的APP使用利用关联挖掘进行了拓展研究。通过关联挖掘并结合电信业业务规则筛选了 101条关联规则,对客户的APP使用进行了关联推荐。文章最后,对全文的主要工作进行了总结,并结合本文存在的不足,对后续的研究进行了展望。
[Abstract]:At present, the competition in the telecommunications industry is becoming more and more fierce, and with the popularity of smart phones and Internet applications, Some network chat tools such as WeChat QQ and other communication software have a certain impact on the traditional business of telecommunications such as SMS and voice services so it is particularly important to segment customers and identify potential customers of value. This paper broadens the index system of customer segmentation with a single index such as customer value, and finally establishes a set of index system suitable to the actual situation by combining the changes brought to enterprises and customers by the deep application of the Internet. In establishing the index system of customer segmentation, this paper integrates the idea of data mining into it. Firstly, based on a large number of raw data available, the paper preprocesses the data, and then establishes the subdivision index system of Internet customers and non-online customers according to the industry characteristics of telecommunications industry. For non-online customers, the index system mainly includes the traditional characteristics of customers, such as telephone, consumption and so on, while for online customers, it mainly includes the usage of customers' online behavior such as traffic, and the browsing behavior of app into the index system. In addition, this paper analyzes the customer characteristics of telecom industry from a variety of angles. In the process of analysis, the online customer and the non-online customer are separated, and the characteristics are analyzed separately. It not only includes the traditional statistical characteristics, such as age, gender, package, terminal, telephone behavior, etc., but also analyzes the application behavior of Internet customers from many aspects, and visualizes the results. In order to overcome the shortcomings of Kohonen SOM algorithm and K-Means algorithm, this paper combines Kohonen SOM algorithm with K-Means algorithm, and establishes Kohonen SOM K-Means clustering analysis model. Kohonen SOM clustering model. Using the results as the initial input of K-Means clustering, the online customers were subdivided into five categories: "ordinary person", "Social King", "Reading fan", "Life Control" and "shopaholic". They are: "inactive customer group", "long distance nocturnal active customer group", "low end active customer group", "high voice affinity network customer group", "higher consumption local customer group" and "high consumption roaming customer group". According to the customer characteristics of different customer groups, the strategies and suggestions for precision marketing of the telecom operation company are put forward. In addition, based on the clustering results, the application mining of high value customers is extended. Through association mining and combining with telecom business rules, 101 association rules are screened, and the application of application is recommended. Finally, the main work of this paper is summarized, and the future research is prospected.
【学位授予单位】：浙江工商大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F626;F274

【参考文献】