下载文档原格式
基于数据挖掘技术的电信业客户流失管理框架
牛琨;张舒博
【期刊名称】《电信工程技术与标准化》
【年(卷),期】2006(19)1
【摘要】本文首先阐述了客户资源的重要性,然后对客户流失这一当今运营商普遍遭遇的难题进行了分析.接着从理论上指出了过去客户保持手段的弊端,分析了防止客户流失的新思路和途径,并总结和归纳了电信业以客户流失管理为主题的数据挖掘项目的几个关键点.在此基础上,提出了基于数据挖掘的客户流失解决方案框架.【总页数】3页(P67-69)
【作者】牛琨;张舒博
【作者单位】北京邮电大学计算机科学与技术学院,北京,100876;中国电信北京研究院决策研究部,北京,100035
【正文语种】中文
【中图分类】F6
【相关文献】
1.基于数据挖掘技术的电信业客户流失分析 [J], 水静;米红娟
2.基于决策树的电信业客户流失分析 [J], 陈捷
3.基于数据挖掘的电信业客户流失分析 [J], 丁红; 陈京民
4.基于数据挖掘的电信业客户流失分析 [J], 丁红; 陈京民
5.基于CART二叉决策树的电信业客户流失的模型构建与控制 [J], 郝梅
因版权原因，仅展示原文概要，查看原文内容请购买。
基于大数据的电信用户流失分析电信用户流失是电信运营商面临的一个重要问题。
随着互联网的普及和竞争的加剧，用户的选择权越来越强，电信运营商需要通过客户流失分析来了解用户的离网原因，并采取相应的措施来减少流失率。
而基于大数据的分析方法可以帮助电信运营商更加全面准确地了解用户离网的原因和规律。
一、大数据在电信用户流失分析中的价值电信运营商的数据规模庞大，包含了用户行为、消费金额、使用时长、网络质量等各个方面的信息。
这些数据量庞大且复杂，传统的数据分析方法已经无法胜任。
而基于大数据的分析方法可以处理海量数据，挖掘出隐藏在数据背后的规律和关联关系。
大数据分析在电信用户流失分析中的价值体现在以下几个方面：1. 精准分析用户特征：通过大数据分析，可以深入了解用户的兴趣、消费偏好、使用习惯等特征。
基于这些特征，可以对用户进行分类，从而为用户提供个性化的服务和推荐，增强用户粘性，减少用户流失。
2. 发现用户流失原因：通过对大数据的挖掘和分析，可以发现用户离网的原因。
比如，通过分析用户的使用行为和网络质量数据，可以发现是否有频繁掉线或网络不稳定的问题，从而及时采取措施改善网络质量，减少用户流失。
3. 预测用户流失趋势：通过建立用户流失预测模型，可以预测用户流失的可能性。
基于这些预测结果，电信运营商可以有针对性地采取措施，提前留住有流失倾向的用户。
4. 监测竞争对手状况：通过对竞争对手的大数据分析，可以了解竞争对手的用户流失情况。
基于这些数据，电信运营商可以及时调整自己的战略，提高竞争力，减少用户流失。
二、基于大数据的电信用户流失分析方法基于大数据的电信用户流失分析方法主要包括数据采集、数据清洗、数据挖掘和流失原因分析四个步骤。
1. 数据采集：首先，需要收集用户的相关数据，包括用户个人信息、消费情况、使用情况、网络质量等。
这些数据可以通过电话清单、短信记录、网络日志、用户调查等方式获取。
2. 数据清洗：由于数据量庞大且来源多样，收集到的数据中难免包含错误和冗余信息。
电信行业大数据分析为电信运营商提供精准营销方案随着互联网的发展和智能设备的普及，电信行业的竞争变得日益激烈。
为了在竞争激烈的市场中保持竞争力，电信运营商需要借助大数据分析技术，为其提供精准的营销方案。
本文将探讨电信行业大数据分析在提供精准营销方案方面的应用和优势。
一、大数据分析在电信行业的应用1. 用户画像分析分析用户画像是电信运营商提供精准营销方案的基础。
通过收集用户的个人信息、通信习惯、消费行为等数据，并结合社交媒体分析、行为分析等多维度数据，可以建立用户的全面画像。
通过对用户画像的分析，电信运营商可以了解用户的需求和偏好，以便更加精准地进行营销推送。
例如，对于观影爱好者的用户可以推送优惠的视频流量套餐，对于商务用户可以推送高速稳定的网络服务。
2. 用户行为分析通过大数据分析用户行为，电信运营商可以获取用户的通信模式、使用场景、使用习惯等信息。
根据用户行为的分析结果，可以针对性地推出定制化产品和服务。
比如，通过分析通讯录联系人的地域分布，可以推测用户的地理位置，从而提供与当地相关的增值业务，比如周边商家特惠推广服务等。
3. 营销效果分析利用大数据技术，电信运营商可以对各类市场活动和推广策略进行数据监测和分析，评估其对用户购买决策的影响。
通过对不同的广告、促销活动的效果进行分析，运营商可以了解哪些策略获得了较好的销售结果，从而优化和调整营销策略，提高市场活动的效果。
二、电信行业大数据分析的优势1. 个性化营销通过大数据分析，电信运营商可以将用户划分为不同的细分群体，对每个群体提供个性化的产品和服务。
这样可以提高用户的满意度和忠诚度，提高销售转化率。
2. 预测用户需求通过对用户行为数据的分析，电信运营商可以预测用户的需求变化趋势。
通过提前调整产品和服务策略，运营商可以更好地满足用户的需求，避免错失商机。
3. 风险控制大数据分析还可以帮助电信运营商进行风险评估和控制。
通过对用户的消费模式和行为进行分析，可以识别潜在的违规行为或欺诈行为，提前采取措施防范风险。
基于数据挖掘的电信运营商用户流失预测模型研究电信运营商是现代社会不可或缺的服务提供者，随着竞争的日益激烈，用户流失成为一个引起广泛关注的问题。
针对这一问题，本文将通过数据挖掘技术构建一个电信运营商用户流失预测模型，并进行相关模型研究。
1. 引言用户流失对电信运营商来说是一个非常重要的指标，因为新用户的获取成本远高于旧用户的保留成本。
通过分析和预测用户流失，电信运营商可以采取一系列的措施来留住用户并提高用户满意度。
2. 数据预处理首先，需要对原始数据进行预处理，清洗掉缺失值和异常值，对数据进行标准化处理，以便更好地进行分析和挖掘。
此外，还需要将数据集划分为训练集和测试集，以便验证模型的准确性和性能。
3. 特征选择在建立用户流失预测模型之前，需要首先选择合适的特征。
常用的特征包括用户基本信息（如性别、年龄、地区等）、用户消费情况（如月消费金额、通话时长、短信发送量等）以及用户满意度等。
通过分析各个特征与用户流失之间的相关性，选择具有预测能力的特征子集。
4. 模型建立本文采用了常用的分类算法来构建电信运营商用户流失预测模型，包括逻辑回归、决策树、支持向量机和随机森林等。
通过对比不同模型的准确性和性能指标，选择最优模型进行后续分析。
5. 模型评估及优化为了评估模型的准确性和性能，可以使用交叉验证、ROC曲线和AUC等指标。
同时，可以对模型进行优化，如调整模型参数、增加特征数量等，以提高预测精度和泛化能力。
6. 结果分析与应用通过对预测结果的分析，可以了解到用户流失的主要原因、影响用户流失的关键因素等。
基于这些分析结果，电信运营商可以制定相应的营销策略，例如给予流失风险较高的用户更多的优惠活动、加强对用户的关怀和沟通等，从而有效降低用户流失率。
7. 研究总结与展望本文通过数据挖掘技术构建了一个电信运营商用户流失预测模型，并进行了相关模型研究。
研究结果显示，通过合理选择特征和优化模型，可以有效地预测用户流失，并采取相应措施来提高用户的满意度和忠诚度。
电信行业的用户流失预测电信行业是一个竞争激烈且充满挑战的行业，用户流失一直是企业关注的焦点。
准确地预测用户流失可以帮助电信公司及时采取措施，提高客户留存率，降低业务成本。
本文将介绍电信行业用户流失预测的方法和应用。
一、用户流失的原因分析用户流失是电信行业常见的问题之一，了解用户流失的原因对于预测和防止流失至关重要。
用户流失的原因可以分为两大类：内外因素。
内因素包括用户满意度、服务质量、产品价格以及竞争对手的优势等。
用户如果对产品或服务不满意，或者竞争对手提供更具吸引力的优惠政策，用户就有可能选择流失。
外因素则包括用户的生活变化、迁居、工作变动等。
这些因素会直接或间接影响用户对电信服务的需求和选择。
了解用户流失的原因可以有针对性地制定预防措施，有效降低用户流失率。
二、预测用户流失的方法为了准确预测用户流失并采取相应的措施，电信公司可以结合数据分析和机器学习等技术手段进行用户流失预测。
1. 数据分析首先，电信公司需要收集并整理用户的历史数据，包括用户的基本信息、使用习惯、消费行为等等。
这些数据可以通过用户登记、账单记录等方式获取。
接下来，通过对历史数据的统计分析，可以发现用户流失的规律和潜在的影响因素。
例如，通过分析用户退订时的共同特征，找出可能导致用户流失的主要因素。
2. 机器学习算法除了数据分析，电信公司还可以利用机器学习算法来提高用户流失预测的准确性。
机器学习是通过训练模型并使用其对新数据进行预测的过程。
电信公司可以使用监督学习算法，根据已知的用户流失情况和相关特征，训练一个预测模型。
然后，使用该模型对新加入或老用户进行预测，判断其是否有流失的可能性。
常用的机器学习算法包括决策树、支持向量机、逻辑回归等。
根据数据的特点和问题的需求，选择适合的机器学习算法进行用户流失预测。
三、用户流失预测的应用用户流失预测的结果可以为电信公司提供宝贵的参考，帮助其制定相应的营销策略和措施，降低用户流失率，提高客户留存率。
大数据对电信运营商流失用户挽回的应用研究随着互联网的高速发展，人们对通信服务的需求也越来越高。
然而，在电信运营商的经营过程中，流失用户是一个常见的问题，也是一个需要重视并解决的挑战。
大数据技术的应用正逐渐成为解决这一问题的利器。
本文将从大数据对电信运营商流失用户挽回的应用研究角度出发，探讨大数据技术在流失用户挽回中的意义以及具体的应用方法。
首先，大数据对于电信运营商来说，是获取用户行为、需求和偏好的重要渠道。
通过收集和分析大量的用户数据，电信运营商可以了解用户的消费习惯、通信需求以及对不同服务的满意度。
这些数据有助于电信运营商精确地掌握用户画像，准确地判断哪些用户正在流失，进而采取相应的措施挽回流失用户。
其次，大数据技术可以通过提供个性化的服务，来增加流失用户的粘性。
个性化服务是电信运营商挽回流失用户的重要手段之一。
大数据技术可以根据用户的偏好和需求，对不同用户进行个性化的推荐和定制服务。
通过准确地分析用户的使用行为和消费模式，电信运营商可以为每个用户提供最优化的服务，从而增加用户的满意度和忠诚度，降低流失率。
第三，大数据技术还可以帮助电信运营商进行流失用户的预测和预警。
通过对大量的历史数据进行分析，电信运营商可以建立流失用户的模型，预测用户的流失概率和时间。
一旦发现有用户有较大的流失可能性，电信运营商可以及时采取相应的措施，例如提供个性化优惠、推送相关的促销活动等，以阻止用户流失，并通过精准的营销活动来挽回用户。
此外，大数据技术还可以帮助电信运营商进行用户群体的细分和定位。
通过对用户的行为数据进行分析，电信运营商可以将用户划分为不同的群体，并了解每个群体的特征和需求。
这将为电信运营商提供准确的用户画像，有针对性地进行流失用户的挽回活动。
不同的用户群体可以采取不同的挽回策略，从而提高挽回效果。
最后，大数据技术还可以通过数据的可视化呈现，为电信运营商的决策提供支持。
通过使用数据可视化技术，电信运营商可以将复杂的数据信息转化为直观的图表和报表，帮助管理人员更好地理解数据，并做出科学合理的决策。
基于数据挖掘的电信用户价值判断与预测电信行业在数字化时代面临着巨大的机遇和挑战。
如何利用大数据技术来进行电信用户价值的判断与预测，成为了电信运营商亟需解决的问题。
本文将基于数据挖掘技术，探讨电信用户价值的判断与预测。
首先，电信用户价值的判断是了解和评估用户对运营商的贡献程度和潜在价值的重要手段。
通过对用户行为、消费习惯等数据进行挖掘和分析，可以了解用户的需求和偏好，从而判断用户对电信运营商的贡献度。
例如，通过挖掘用户的通话记录、短信记录和上网记录，可以了解用户的通讯频率和上网偏好；通过挖掘用户的消费记录，可以了解用户的消费水平和偏好等。
电信运营商可以根据用户的特点和需求，有针对性地进行产品推荐、促销活动等，提高用户的满意度和忠诚度。
其次，电信用户价值的预测是根据用户的历史行为和消费特征，预测用户未来的行为和消费趋势。
通过利用数据挖掘技术，如分类、聚类、关联分析等，可以将用户分为不同的群体，并预测用户的行为和消费趋势。
例如，可以根据用户的消费行为和消费频率，预测用户未来是否会继续使用某项业务或购买某种产品；可以根据用户的上网行为，预测用户未来的上网时间和流量需求。
电信运营商可以根据用户的预测结果，调整产品策略和市场推广策略，提前满足用户的需求，提高用户的满意度和忠诚度。
数据挖掘在电信用户价值判断与预测中发挥着重要作用。
数据挖掘是一种通过自动发现隐藏在大数据中的模式和规律，提供有价值信息和知识的技术。
在电信行业中，大量的用户数据积累了用户的行为、消费、信用等方面的信息。
通过数据挖掘技术，可以从海量数据中提取出有用的信息和知识，为电信运营商提供决策支持和市场预测。
数据挖掘的核心任务之一是分类。
分类是将数据集划分为不同的类别，即根据已有的标签信息来预测新数据的类别。
在电信用户价值判断中，可以利用分类算法将用户划分为不同的类别，如高价值用户、中等价值用户和低价值用户。
通过对不同类别用户的特点和行为进行分析，电信运营商可以制定不同的增值服务和营销策略，提高用户的满意度和忠诚度。
基于大数据分析的电信运营商用户流失预测研究电信运营商用户流失是一个长期以来一直困扰着电信行业的问题。
用户流失的发生不仅导致运营商的收入减少，还影响了用户体验和品牌形象。
因此，预测用户流失并采取相应的措施来挽留用户成为了电信运营商的重要任务之一。
近年来，随着大数据技术的发展，电信运营商开始广泛应用大数据分析来预测用户流失，以提高用户满意度和保持竞争力。
本文将基于大数据分析的电信运营商用户流失预测进行研究，通过对用户数据的挖掘和分析，提出一种预测模型，以帮助电信运营商准确预测用户流失，及时采取措施挽留用户。
首先，我们需要收集和整理大量的用户数据，包括用户的基本信息、消费行为、网络活动等。
这些数据可以通过用户注册信息、用户通话记录、网络浏览记录等渠道获取。
同时，为了数据的准确性和完整性，我们还可以结合其他数据源，如第三方数据和社交媒体数据。
收集到的数据将作为预测模型的输入变量。
接下来，我们需要对收集到的用户数据进行清洗和处理，以消除数据中的噪声和异常值。
清洗后的数据将用于构建预测模型。
在预测模型的构建过程中，我们可以使用多种方法，如决策树、逻辑回归、支持向量机和人工神经网络等。
在模型构建之前，我们需要对数据进行特征工程。
特征工程是指选择和构建与用户流失相关的特征变量，在模型中起到解释和预测的作用。
常用的特征工程方法包括特征选择、特征变换和特征创造。
通过特征工程，我们可以从大量的用户数据中筛选出对用户流失有影响的关键特征。
在模型构建过程中，我们还需要划分训练集和测试集。
训练集用于模型的训练和参数调优，测试集用于模型性能的评估。
为了提高模型的准确性和稳定性，我们可以采用交叉验证的方法来进行模型的选择和评估。
完成模型的训练后，我们可以使用模型进行用户流失的预测。
预测结果可以帮助电信运营商针对潜在流失用户采取个性化的挽留策略，如降价促销、赠送礼品、提供优质客户服务等。
同时，我们还可以通过对预测结果的分析来发现用户流失的原因和规律，从而进一步优化产品和服务，提高用户满意度和黏性。
电信运营的大数据分析洞察用户行为和市场趋势的重要工具在数字化时代，随着互联网的蓬勃发展，通信技术也得到了极大的提升。
作为通信行业的重要组成部分，电信运营商积累了大量的用户数据。
这些海量的数据所蕴含的信息价值是不容小觑的。
通过对这些数据进行深入分析，电信运营商可以洞察用户的行为特点和市场趋势，并据此制定相应的运营策略。
因此，大数据分析被视为电信运营的重要工具，具有重要的意义。
一、用户行为洞察通过对用户数据的挖掘和分析，电信运营商可以了解用户的使用习惯、偏好以及需求。
这些信息对于电信运营商来说十分关键，有助于他们提供更加个性化和精准的服务。
例如，运营商可以通过分析用户的通信消费行为和数据使用情况，为用户量身定制套餐，从而提高用户的满意度和忠诚度。
同时，通过对用户行为的洞察，运营商还可以发现潜在的付费意愿，为增加收入提供有力支持。
二、市场趋势洞察大数据分析不仅可以洞察用户行为，还可以帮助电信运营商洞察市场趋势。
通过分析用户在通信市场上的表现，运营商可以了解不同市场细分的发展趋势和用户需求变化。
这对电信运营商来说具有重要的指导意义。
比如，在竞争激烈的移动通信市场，通过对用户数据的分析，运营商可以发现新的业务增长点和市场机会，并进行相应的产品和服务创新。
同时，通过对市场趋势的洞察，电信运营商可以优化资源配置，提高经营效率。
三、大数据分析的技术手段要实现对用户行为和市场趋势的洞察，电信运营商需要借助大数据分析的技术手段。
大数据分析通常包括数据采集、数据清洗、数据存储、数据挖掘和数据可视化等环节。
首先，在数据采集环节，电信运营商可以通过不同的渠道和方式搜集用户数据，包括通信日志、上网记录、用户调查等。
这些数据通常以结构化和非结构化的形式存在。
其次，在数据清洗环节，电信运营商需要对采集到的数据进行清洗和整理，去除重复或无效的数据，保证数据的质量和准确性。
然后，在数据存储环节，电信运营商需要建立起相应的数据存储系统，以便对海量的数据进行存储和管理。
I.J. Education and Management Engineering, 2017, 1, 36-45Published Online January 2017 in MECS ()DOI: 10.5815/ijeme.2017.01.04Available online at /ijemeFramework for Targeting High Value Customers and Potential Churn Customers in Telecom using Big Data AnalyticsInderpreet Singh a, Sukhpal Singh ba Engineer- Product Development, Mahindra Comviva, Gurgaon-122001, Haryana, Indiab Computer Science and Engineering Department, Thapar University-147001, Patiala, Punjab, IndiaAbstractSince the more importance is played on customer’s behavior in today’s business market, telecom companies are not only focusing on customer’s profitability to increase their mar ket share but also on their potential churn customers who could terminate the relation with the company in near future. Big data promises to promote growth and increase efficiency and profitability across the entire telecom value chain. This paper presents a framework for targeting high value customers and potential churn customers. Firstly, customers are segmented on basis of RFM (Recency-Frequency-Monetary) analysis and finally customers in each segment are targeted by various offers on basis of their similar characteristics.Index Terms: Customer Segmentation, Telecom, Big Data Analytics, RFM Analysis, Data Mining, Customer Value.© 2017 Published by MECS Publisher. Selection and/or peer review under responsibility of the Research Association of Modern Education and Computer Science.1.IntroductionCustomer equity plays an important role in today’s environment. Many organizations are now focusing on analyzing the big data related to customer for improving profitability and market share [3]. Due to rise in complexity and competition in current business scenario, organizations need to develop activities that are innovative to capture the needs of customer and retention [11].Service providers that are trying to compete in the cutthroat world of telecom services –where more and more subscribers rely on over-the-top players as providers of value-added services – are focused on increasing revenues, reducing churn, reducing opex, and enhancing the customer experience [18].Big data is phrase for data sets that are so huge and complicated, making it difficult to process by traditional data processing applications. The size of these data sets is increasing day by day making it harder to store, analyze and unleash its full potential [17].* Corresponding author: Tel.: +919855500543E-mail address: inderpreet.singh993@During recent years, smart phones have become our universal and most itinerant computing resource. The transformations in smart phones have made it prone to wide range of applications [1]. At present every operator is looking for various ways of leveraging data to increase the revenues and market share [19] [20]. But they want to take care of unstructured data along with the structured data to understand the consumer behavior. So that they could gain competitive edge in the market. Gartner forecasted mobile data growth to reach at 52 million terabytes in 2015 i.e. an increase of 59 percent from the year 2014 [2]. More than 600 million tweets are produced on Twitter on each day [21]. About 800 million members of Facebook are active on daily basis. Imagine the data produced by their posts, comments or chats [23].Most operators run analysis on their internal data to raise the efficiency of their networks, do the customer segmentation and get profit [22]. But the main challenge that they face is how to combine the enormous amount of information to increase the profits across entire industry, from customer satisfaction to product development and sales etc. [4].The objective of this paper is to provide a framework to telecom companies for getting competitive edge in the market by identifying not only their high value customers but also potential churn customers. In this paper, firstly we group the customers in various segments on basis of structured and unstructured data sets. Then finally, we target the customers in high value and potential churn segment. The customers in different segments are targeted on basis of different offers made specifically for them.The paper is organized as follows. Section 2 presents related work. The various types of data sets needed and how to tap them is explained in the section 3, 4 and 6. The section 7 explains what the RFM (Recency-Frequency-Monetary) analysis is and it is done on the collected data as mentioned in section 8.Then, finally section 9 explains the methods to target the high value and potential churn customers found through the RFM analysis done in previous sections. Section 10 presents conclusion and future scope of the paper.2.Related WorkIn [12], authors discussed about the system which selects eligible users for predefined promotion and then analyzing the Caller Detail Records, while rapidly escalating subscriber base. In our paper, we are analyzing customers on various data sets and then targeting the high value and potential churn segment of the customers. In [13], authors explains the big data platform for processing and visualization of large data files for telecom companies. In addition to this they also have designed Distributed File Descriptor (DFD) for passing information between the modules of platform i.e. Domain Specific Language (DSL), Parallel Processing/Analyzing Platform for Big Data and an Integrated Result Viewer.While in [14], authors proposed the practical application of data mining in customer segmentation under the data environment of telecommunication industry. They have done on basis of cluster analysis. In [15], authors addresses the Social position of each customer in a network and Equivalence approaches. Social position which is evaluated by finding the centrality of a node identified through a number of connections. While regular equivalence analysis seeks to classify customers as churners and non-churners based on the patterns. Finally in [16], authors explains how both decision tree and neural network techniques can deliver accurate churn prediction models by using various data sets.3.Structured And Unstructured DataBy Structured data we mean data that can be allocated to particular fields and can be handled by computing resources [4]. It has great degree of organization, so it can be inserted into relational databases very smoothly (Fig. 1). The perfection of its structure makes compilation easy and fast task. Whereas by unstructured data we mean the data that has irregularities and ambiguities making it difficult to handle by traditional resources. Unstructured data doesn’t fits into data models defined prior. The significant amount of data comes from unstructured data category and it’s the tough task for any organization to make use of this unstructured dataalong with the structured data to get better insights of consumers in the market (Fig. 2).Fig.1. Structured Data and Unstructured Data [5]Fig.2. Structured Data and Unstructured Data [5]4.Various Types Of Data NeededTo make profile of each and every consumer telecom companies need to gather various types of data, so that segmentation on consumer market can be performed efficiently. And based on this segmentation we can target the customers with various types of offers suitable for his or her needs. The set of structured data that we need to make costumer profile is as mentioned in the below table (Table 1.),Table 1. Set of Structured DataAlong with the set of structured data we also need set of unstructured data to make customer profile. Thus, helping us in making creative insights of consumer market. So, the set of unstructured data that we need to make costumer profile is as mentioned in Table 2.Table 2. Set of Unstructured Data5.Need To Address 3V ChallengeDoug Laney [6] was first to characterize 3V’s of big data. These are variety, volume and velocity. In other words, high variety, volume and velocity are the parameters of data that characterize it as a big data (Fig. 3). The size of data set can range from few petabytes to many Exabyte’s.Fig.3. 3V’s of Big Data5.1.VarietyThe first characteristic of data is the growing need to use variety of data, i.e. now data represent number of data domains and types. And given the large range of data types it is difficult to integrate the data for its use in analysis. Like we have to integrate both the structured and unstructured data on basis of which we will be analyzing the consumer market.5.2.VolumeV olume is most associated with the big data. 90% of world’s data ever generated was generated in past two years [7]. If this trend continues then by 2020, data generated will be three times than the data generated till 2011. This huge volume of data presents biggest challenge to conventional information structures. Hence companies find it difficult to store this much large volume of data.5.3.VelocityThe speed at which data is generated, stored, analyzed and visualized is represented by velocity of data. In this era, data is being created at the real time and the problem faced by the companies is to give real time solutions to this data.6.How To Get Data- Structured And UnstructuredThe general problem faced by the companies is to tap the unstructured data. While the structured data can be collected and stored easily in well-defined data model e.g. RDBMS, unstructured data doesn’t has any particular format. In case of telecom operators, the structured data can be easily collected from the customer information form, which the customer fills when he gets registered. And the structured data like number of minutes used, amount of internet data used can be easily collected from the transaction resources which operators usually have in form of databases.Unstructured data like geo spatial information can be taken from two sources. The first one is by satellite i.e. GPS (Global Positioning System).GPS provides precise location and time information in all types of weather conditions. And other one is by Wi-Fi Positioning System as within buildings GPS doesn’t work well because of obstructions in the line of sight with the satellites [10]. Wi-Fi Positioning System is used where GPS is inadequate. So, by getting information from two sources operator will be precise in determining the exact location of customer.Social media data could be fetched through Twitter API’s or customer post on Faceboo k page of telecom operator etc.7.RFM AnalysisRFM analysis guide the organization to analyze the customer value on basis of which organization can predict which customers are likely to respond as well as that will not respond to their offer. Each customer is evaluated on basis of recency, frequency and monetary value in RFM analysis. In simple words, RFM is technique of marketing analysis which is used to examine which customers are best ones by examining how recent customer has made purchase (Recency), how much often they purchase (Frequency) and how much he or she spends (Moneta ry). It is based on market law that ―80% of business comes from 20% of your customers‖. The three inputs that make the RFM scores are-∙Recency- What was the time when the customer had made purchase with your organization. A customer who recently contacts your organization is more likely to accept another interaction.∙Frequency-How often does this customer make purchases with your organization. It can predict the schedule of purchases which will be made by customer in future.∙Monetary-How much total amount customer spend on your organization products in the given time period.Customers are then given the ratings on the basis of these three input parameters. The higher the RFM analysis score, the higher is the customer value for your organization. Now telecom operators can generate RFM scores individually for the various parameters. After that, they can use these scores not only to retain the higher value customers but also to gain the potential churn customers back.8.Calculating RFM ScoreTo calculate RFM scores we have divided the three input parameters to five subcategories. Initially we havefound the scores of recency and frequency inputs and then combine these scores with the third input i.e. monetary to get overall RFM analysis score.The various subcategories under the input variables are:8.1.Recency (Number of days)Here we have divided the customers into different recency categories (Table 3.) according to the number of days since the last transaction. With R1 as customers who have not made transaction with the organization since 121 or more days and R5 as customers who have made transaction with the organization within 30 days or less etc.Table 3. Evaluation Criteria for Recency Score8.2.Frequency (Number of times)In this we have divided the customers into different frequency categories (Table 4.). According to the total number of times he or she has made transaction with the organization within last 6 months. F1 as customers who have made less than or equal to 2 transactions in last 6 months and F5 as customers who have made greater than or equal to 11 transactions in last 6 months.Table 4. Evaluation Criteria for Frequency Score8.3.Monetary(In INR)In this we have divided the customers into different monetary categories (Table 5.) according to the total amount that he or she has paid for transactions with the organization within last 6 months. F1 as customers who have paid less than or equal 100 (INR) transactions in last 6 months and F5 as customers who have paid greater than or equal 10000 (INR) transactions in last 6 months.Here, INR- Indian Rupee.Table 5. Evaluation Criteria for Monetary ScoreCalculating RF scores (Table 6.),Table 6. Evaluation Creteria for RF(Recency-Frequency) ScoreOn the basis of RF scores calculated above, we are calculating the RFM scores (Table 7.),Table 7. Evaluation Criteria for RFM Score9. Customer Segmentation And Targeting Them9.1. Customer SegmentationCustomer segmentation involves dividing a customer base into the different subsets where customers of specific subset have similar interest and spending habits [8]. On basis of RFM scores as calculated above we can divide the customers into two segments.High Value CustomersT hese are the clients on which operator’s profitability depends. Without them operator will lose its share inthe telecom market and competitive advantage. These clients are paid proper care and attention from the operator.Potential churn CustomersPotential churn customers are those customers that can cease or end their relation with the operator. The total cost (1) associated with these potential churn customers if they cease their relationship isTotal Cost (Customer Churn) = Lost Revenue + Marketing Cost (1) Lost revenue is the revenue that these customers could generate, had they not ceased their relationship with operator. Marketing cost is the cost associated with the replacement of these customers with new ones [9].9.2. Targeting the customersBy calculating the RFM score, individually for the case of total time spent on calls and internet data pack’s total data. We have recognized the high value customers as well as potential churn customers. But in today’s world most of the persons get pack of telecom operator services which include both call rate and internet data amount. Sudden exceptions to this are the persons who only use operator services for Internet or Calls not both of them. So, to target customers (Table 8.) on basis of these things we need to look at both the RFM scores of total time of calls and amount of internet data.Table 8. Criteria to Target CustomersHigh- Customers using large internet data and spending much time on calls.Mid- Customers using both mediocre internet data and spending time on calls.Low- Customers using less internet data and spending less on calls.High users can be targeted by the use of loyalty points and personalized offers specially designed for them, as they are the key for the telecom operator’s competitive edge in the market, for the mid or mediocre users operators can give offers based on combination of both call rates and data pack and for the low users operators can give offers that tend to cause these users to use more services provided by the operators e.g. free local calls to same telecom operator mobile numbers at late night etc.For potential churn customers i.e. ―Low‖ mentioned in Table 8, combine RFM score with the unstructured data like social media data and call center feedback data to predict them very accurately. Potential churn customers should also be considered seriously by the telecom operators due to revenue and marketing cost associated with both of them. So, telecom operator must give offers like free talk time and data pack for specific time period e.g. 200MB data for 3 days, in order to retain them.10.Conclusion And Future WorkThis paper discussed the approach that how could telecom operators can collect the data, analyze it and segment their customers to gain competitive advantage in the market. Customer segmentation not only helps the telecom operators to know their high value customers on which their profitability and market share depends, but also to recognize the potential churn customers who are likely to cease their relationship with the operator. It also discusses the methods to target these high value customers and potential churn customers.Big data used effectively has the potential to revolutionize the way telecom operators build, run and market their services. The presented framework can be used to help the telecom operators in analytics related practices. Some open issues that need special attention in future research are,∙Approach to integrate above framework with the real time analytics to target customer on basis of location, time etc. in real time.∙To make versatile model for telecom sector that can be used to target customers on basis of different conditional inputs.References[1]Goggin, G. Cell phone culture: Mobile technology in everyday life. Routledge, (2012).[2]―Gartner Forecasts 59 Percent Mobile Data Growth Worldwide In 2015". 2016. ./newsroom/id/3098617.[3]Acker, O., Blockus, A. and Pötscher, F. "Benefiting from big data: A new approach for the telecomindustry." Strategy&, Analysis Report (2013).[4]Baars, H. and Kemper, H.G. "Management support with structured and unstructured data—an integratedbusiness intelligence framework." Information Systems Management 25, no. 2 (2008): pp 132-148.[5]The Executive’s Guide to Big Data & Apache Hadoop written by Robert D. Schneider; Page9.[6]Laney, D. "3D data management: Controlling data volume, velocity and variety." META GroupResearch Note 6 (2001): pp 70.[7]Data, Big. "for better or worse: 90% of world’s data generated over last two years." SCIENCE DAILY,May 22 (2013).[8]"What Is Customer Segmentation? - Definition from ." SearchSalesforce. Accessed August20, 2016. /definition/customer-segmentation. [9]"Customer Churn Software: Prediction, Prevention & Analysis | Optimove". 2016. Optimove./learning-center/customer-churn-prediction-and-prevention.[10]Li, B., Mumford, P., Dempster, A.G. and Rizos, C. "Secure User Plan Location (SUPL): concept andperformance," GPS Solutions, vol. 14, pp. 153, (2010).[11]Chuang, H.M. and Shen, C.C. "A study on the applications of data mining techniques to enhancecustomer lifetime value—based on the department store industry." In 2008 International Conference on Machine Learning and Cybernetics, vol. 1, pp. 168-173. IEEE, (2008).[12]Jayawardhana, P., Perera, D., Kumara, A. and Paranawithana, A "Kanthaka: Big Data Caller DetailRecord (CDR) Analyzer for Near Real Time Telecom Promotions." In 2013 4th International Conference on Intelligent Systems, Modelling and Simulation, pp. 534-538. IEEE, (2013).[13]Şenbalcı, C., Altuntaş, S., Bozkus, Z. and Arsan, T. "Big data platform development with a domainspecific language for telecom industries." In 2013 High Capacity Optical Networks and Emerging/Enabling Technologies, pp. 116-120. IEEE, (2013).[14]Shi, J.Y. and Li, L.L. "The Research of Data Mining in Telecom Data Warehouse." In System Science,Engineering Design and Manufacturing Informatization (ICSEM), 2010 International Conference on, vol. 2, pp. 239-242. IEEE, (2010).[15]Shobha, G. "Social network classifier for churn prediction in telecom data." In Advanced Computingand Communication Systems (ICACCS), 2013 International Conference on, pp. 1-7. IEEE, (2013). [16]Hung, S.Y., Yen, D.C. and Wang. "Applying data mining to telecom churn management." ExpertSystems with Applications 31, no. 3 (2006): pp. 515-524.[17]Singh, S. and Chana, I. ―EARTH: Energy-aware autonomic resource scheduling in cloudcomputing.‖Journal of Intelligent & Fuzzy Systems Vol. 30, no. 3 (2016): pp. 1581-1600[18]Singh, S. and Chana, I. ―Resource provisioning and scheduling in clouds: QoS perspective.‖TheJournal of Supercomputing Vol. 72, no. 3 (2016): pp. 926-960.[19]Singh, S. and Chana, I. "QoS-aware autonomic resource management in cloud computing: a systematicreview." ACM Computing Surveys (CSUR) Vol. 48, no. 3 (2016): pp. 42.[20]Singh, S. and Chana, I. ―QRSF: QoS-aware resource scheduling framework in cloud computing‖, ―TheJournal of Supercomputing‖, [Springer], Vol. 71, no. 1, pp: 241-292, 2015.[21]Ahmed T. "Cloud Computing a Solution for Globalization", International Journal of Education andManagement Engineering (IJEME), Vol.6, No.4, pp.30-38, 2016.DOI: 10.5815/ijeme.2016.04.04. [22]Herrera J. M., González J. V. and Fillad L.L. "Web System Proposal for Control and Monitoring Fleets",International Journal of Education and Management Engineering (IJEME), Vol.6, No.1, pp.1-10, 2016.DOI: 10.5815/ijeme.2016.01.01.[23]Kolo K.D., Adepoju S.A. and Alhassan J.K.,"A Decision Tree Approach for Predicting StudentsAcademic Performance", International Journal of Education and Management Engineering (IJEME), Vol.5, No.5, pp.12-19, 2015.DOI: 10.5815/ijeme.2015.05.02.Authors’ ProfilesInderpreet Singh has obtained the Degree of Bachelor of Engineering in Computer Sciencefrom Thapar University, Patiala. Presently, he is working as an Engineer-Product Developmentin Mahindra Comviva, Gurgaon. He has done certifications in PostgreSQL, Cloud Computingand Python Fundamentals. His research interests include Analytics, Data Mining, PredictiveModeling and Cloud Computing.Sukhpal Singh joined Computer Science and Engineering Department of Thapar University,Patiala, India, in 2016 as Lecturer. Dr. Singh obtained the Degree of Master of Engineering inSoftware Engineering from Thapar University, as well as a Doctoral Degree specialization in―Autonomic Cloud Computing‖ from Thapar University, Dr. Singh received the Gold Medal inMaster of Engineering in Software Engineering. Dr. Singh is a DST Inspire Fellow [2013-2016]and worked as a SRF-Professional on DST Project, Government of India. He has done certifications in Cloud Computing Fundamentals, including Introduction to Cloud Computing and Aneka Platform (US Patented) by ManjraSoft Pty Ltd, Australia and Certification of Rational Software Architect (RSA) by IBM India. His research interests include Software Engineering, Cloud Computing, Operating System and Databases. He has more than 30 research publications in reputed journals and conferences.How to cite this paper: Inderpreet Singh, Sukhpal Singh,"Framework for Targeting High Value Customers and Potential Churn Customers in Telecom using Big Data Analytics", International Journal of Education and Management Engineering(IJEME), Vol.7, No.1, pp.36-45, 2017.DOI: 10.5815/ijeme.2017.01.04。
