电信用户稳定性识别及应用研究

发布时间：2018-01-18 10:42

本文关键词：电信用户稳定性识别及应用研究　出处：《首都经济贸易大学》2017年硕士论文　论文类型：学位论文

【摘要】：世界正走向一个信息化、数据化的时代,科技的进步激发了电信产业的发展,过去几年,电信业成为全球经济增长最快的行业,中国移动、中国联通和中国电信国内三大通讯巨头在国内外纷纷上市,使得电信产业成为推动国民经济发展的主要支柱。但近年来,随着我国电信行业的改革和重组不断推进,电信应用的不断普及,电信市场的竞争日益激烈,基础业务的收入也趋于饱和,ARPU(每用户平均收入)出现明显下滑,用户的忠诚度也在不断降低。当前市场状况下,维持老用户不流失,促进新用户转型成稳定型用户,确定出具有稳定性潜质的新用户,识别出用户中高价值稳定的用户,并根据用户特征细分用户并进行个性化服务,增强用户体验,从而促进收益增长,成为了电信业工作的重中之重。本文基于某运营商某年某月的用户话单行为数据,采用分类、聚类方法,提出用户稳定性识别模型和用户分群精准化营销方式,主要研究内容包括:1.数据的清洗与预处理,包括对数据的统一化处理、去噪声处理、缺失值处理等;2.根据用户生命周期理论,发现入网时长与流失速度的变化规律,通过入网时长来界定用户的稳定性,运用随机森林、adaboost、支持向量机三种方法建立了用户稳定性识别模型,在模型比较后,最后选取随机森林作为用户稳定性识别的模型;3.根据用户的稳定性特征和用户行为属性数据对用户进行混合型聚类分析,将用户聚成了6类,刻画了每类群体的用户画像;4.营销策略与建议,根据每类群体用户的稳定性和显著的行为特征,采取个性化服务,实现精准营销,增强用户体验,从而为运营商创造价值。
[Abstract]:The world is moving towards an era of information and data. The progress of science and technology has stimulated the development of the telecommunications industry. In the past few years, the telecommunications industry has become the fastest growing industry in the world, China Mobile. China Unicom and China Telecom are listed at home and abroad, which makes the telecommunications industry become the main pillar to promote the development of the national economy. But in recent years. With the continuous reform and restructuring of the telecommunications industry in China, telecommunications applications continue to be popularized, the competition in the telecommunications market is becoming increasingly fierce, and the revenue of basic services tends to saturate. ARPU( average income per user) has declined significantly, and the loyalty of users is also declining. Under the current market conditions, maintain the old users do not lose, and promote the new users to become stable users. Identify new users with stability potential, identify users with high value and stability, subdivide users according to the characteristics of users and carry out personalized services, enhance user experience, thereby promoting revenue growth. In this paper, based on a certain operator of a certain year and a month of user behavior data, using classification, clustering method, proposed user stability identification model and user group accurate marketing method. The main research contents include: 1. Data cleaning and preprocessing, including unified data processing, noise removal, missing value processing, etc. 2.According to the theory of user life cycle, this paper finds out the changing law of the time of entering the network and the speed of loss, defines the stability of the user by the time of entering the network, and uses random forest adaboost. Three methods of support vector machine (SVM) are used to establish the user stability identification model. After the model comparison, random forest is selected as the user stability identification model. 3.According to the stability characteristics of users and the user behavior attribute data, the users are clustered into six categories, and the user portraits of each group are depicted. 4. Marketing strategies and suggestions, according to the stability of each group of users and significant behavioral characteristics, the adoption of personalized services to achieve precision marketing, enhance user experience, thereby creating value for operators.
【学位授予单位】：首都经济贸易大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F626

【参考文献】