Modeling churn using customer lifetime value

The definition and modeling of customer loyalty have been central issues in customer relationship management since many years. Recent papers propose solutions to detect customers that are becoming less loyal, also called churners. The churner status is then defined as a function of the volume of commercial transactions. In the context of a Belgian retail financial service company, our first contribution is to redefine the notion of customer loyalty by considering it from a customer-centric viewpoint instead of a product-centric one. We hereby use the customer lifetime value (CLV) defined as the discounted value of future marginal earnings, based on the customer’s activity. Hence, a churner is defined as someone whose CLV, thus the related marginal profit, is decreasing. As a second contribution, the loss incurred by the CLV decrease is used to appraise the cost to misclassify a customer by introducing a new loss function. In the empirical study, we compare the accuracy of various classification techniques commonly used in the domain of churn prediction, including two cost-sensitive classifiers. Our final conclusion is that since profit is what really matters in a commercial environment, standard statistical accuracy measures for prediction need to be revised and a more profit oriented focus may be desirable.     

New insights into churn prediction in the telecommunication sector: A profit driven data mining approach

Customer churn prediction models aim to indicate the customers with the highest propensity to attrite, allowing to improve the efficiency of customer retention campaigns and to reduce the costs associated with churn. Although cost reduction is their prime objective, churn prediction models are typically evaluated using statistically based performance measures, resulting in suboptimal model selection. Therefore, in the first part of this paper, a novel, profit centric performance measure is developed, by calculating the maximum profit that can be generated by including the optimal fraction of customers with the highest predicted probabilities to attrite in a retention campaign. The novel measure selects the optimal model and fraction of customers to include, yielding a significant increase in profits compared to statistical measures.In the second part an extensive benchmarking experiment is conducted, evaluating various classification techniques applied on eleven real-life data sets from telecom operators worldwide by using both the profit centric and statistically based performance measures. The experimental results show that a small number of variables suffices to predict churn with high accuracy, and that oversampling generally does not improve the performance significantly. Finally, a large group of classifiers is found to yield comparable performance.     

Predicting partial customer churn using Markov for discrimination for modeling first purchase sequences

Currently, in order to remain competitive companies are adopting customer centered strategies and consequently customer relationship management is gaining increasing importance. In this context, customer retention deserves particular attention. This paper proposes a model for partial churn detection in the retail grocery sector that includes as a predictor the similarity of the products?? first purchase sequence with churner and non-churner sequences. The sequence of first purchase events is modeled using Markov for discrimination. Two classification techniques are used in the empirical study: logistic regression and random forests. A real sample of approximately 95,000 new customers is analyzed taken from the data warehouse of a European retailing company. The empirical results reveal the relevance of the inclusion of a products?? sequence likelihood in partial churn prediction models, as well as the supremacy of logistic regression when compared with random forests.     

Fuzzy measure of fuzzy events defined by fuzzy integrals

We define the concept of fuzzy measure of a fuzzy event by using a general form of fuzzy integral proposed by Murofushi, called fuzzy t-conorm integral, encompassing previous definitions. Zadeh defined the probability measure of a fuzzy event, and later the possibility measure of fuzzy event. Using a duality property of fuzzy t-conorm integral, we propose a general definition of fuzzy measure of fuzzy events, which is compatible with previous definitions of Zadeh, and possesses all properties of a fuzzy measure, in particular the duality property. Using our definition, we examine the case of decomposable measures and belief functions. A comparison with previous works is provided.     

A hierarchical multiple kernel support vector machine for customer churn prediction using longitudinal behavioral data

The availability of abundant data posts a challenge to integrate static customer data and longitudinal behavioral data to improve performance in customer churn prediction. Usually, longitudinal behavioral data are transformed into static data before being included in a prediction model. In this study, a framework with ensemble techniques is presented for customer churn prediction directly using longitudinal behavioral data. A novel approach called the hierarchical multiple kernel support vector machine (H-MK-SVM) is formulated. A three phase training algorithm for the H-MK-SVM is developed, implemented and tested. The H-MK-SVM constructs a classification function by estimating the coefficients of both static and longitudinal behavioral variables in the training process without transformation of the longitudinal behavioral data. The training process of the H-MK-SVM is also a feature selection and time subsequence selection process because the sparse non-zero coefficients correspond to the variables selected. Computational experiments using three real-world databases were conducted. Computational results using multiple criteria measuring performance show that the H-MK-SVM directly using longitudinal behavioral data performs better than currently available classifiers.     

Tracking customer portfolio composition: A factor analysis approach

In this paper, we address the changing composition of a customer portfolio taking into account actions undertaken by the company to adapt its service offer to market conditions and/or technological innovations. We present a specific methodology to identify clusters of customers in different periods and then compare them over time. The classification process takes into account both qualitative and quantitative aspects of the consumption levels of the services or products offered by the company. The possibility of period‐to‐period variation in the customer portfolio and the service or product offer is also considered, in order to achieve a more realistic scenario. The core of the proposed methodology is related to the family of exploratory factorial and cluster techniques. The customers are classified by using a bicriterial clustering methodology based on ‘tandem’ analysis (multiple factor analysis+cluster analysis of the main factors). The bicriterial approach allows for a compromise between customers' consumption levels (a quantitative criterion) and their consumption/non‐consumption pattern (a qualitative criterion). The evolution of the customer portfolio composition is explored through multiple correspondence analysis. This technique allows visual comparison of the position of different clusters against time and the identification of key changes in customer consumption behavior. The methodology is tested on realistic customer portfolio scenarios for a major telecommunication company. We simulate various scenarios to show the strengths of our proposal. Copyright © 2009 John Wiley & Sons, Ltd.     

ADTreesLogit model for customer churn prediction

In this paper, we propose ADTreesLogit, a model that integrates the advantage of ADTrees model and the logistic regression model, to improve the predictive accuracy and interpretability of existing churn prediction models. We show that the overall predictive accuracy of ADTreesLogit model compares favorably with that of TreeNet®, a model which won the Gold Prize in the 2003 mobile customer churn prediction modeling contest (The Duke/NCR Teradata Churn Modeling Tournament). In fact, ADTreesLogit has better predictive accuracy than TreeNet® on two important observation points.     

On regular duo po-Γ-semigroups

There are two definitions of Γ-semigroups and investigations for both of them in the bibliography. The definition given by Sen in 1981, and the definition given by Sen and Saha in 1986. In this paper we show the way we pass from ordered semigroups to ordered Γ-semigroups no matter which one of the two definitions we use. Moreover we show that, exactly as in ordered semigroups, in many results of ordered Γ-semigroups points do not play any essential role, but the sets, which shows their pointless character. Under the methodology using in this paper, all the results of ordered semigroups can be transferred into ordered Γ-semigroups.     

移动电话客户流失数据挖掘

本文首先回顾了顾客流失的相关文献,然后利用统计分析方法和数据挖掘技术分析了移动电话号码与移动电话型号对客户流失的影响,对分析结果进行了解释,并给出一些营销建议。     

On security properties of all-or-nothing transforms

All-or-nothing transforms have been defined as bijective mappings on all s-tuples over a specified finite alphabet. These mappings are required to satisfy certain “perfect security” conditions specified using entropies of the probability distribution defined on the input s-tuples. Alternatively, purely combinatorial definitions of AONTs have been given, which involve certain kinds of “unbiased arrays”. However, the combinatorial definition makes no reference to probability definitions. In this paper, we examine the security provided by AONTs that satisfy the combinatorial definition. The security of the AONT can depend on the underlying probability distribution of the s-tuples. We show that perfect security is obtained from an AONT if and only if the input s-tuples are equiprobable. However, in the case where the input s-tuples are not equiprobable, we still achieve a weaker security guarantee. We also consider the use of randomized AONTs to provide perfect security for a smaller number of inputs, even when those inputs are not equiprobable.

     

The orienteering problem with stochastic travel and service times

In this paper, we introduce a variant of the orienteering problem in which travel and service times are stochastic. If a delivery commitment is made to a customer and is completed by the end of the day, a reward is received, but if a commitment is made and not completed, a penalty is incurred. This problem reflects the challenges of a company who, on a given day, may have more customers than it can serve. In this paper, we discuss special cases of the problem that we can solve exactly and heuristics for general problem instances. We present computational results for a variety of parameter settings and discuss characteristics of the solution structure.     

On fuzzy cliques in fuzzy networks

A definition of fuzzy clique in social networks is suggested which overcomes five limitations of current definitions. This definition is based on the networks in which the 0–1 strengths, the weighted strengths, and fuzzy strengths are all allowed. The fuzzy distance in such a network is defined. The node‐clique and clique‐clique coefficients are suggested. The core and the periphery of fuzzy cliques are discussed formally. A “cone like” property of the cores is discovered. The network structures are discussed using the new definition. A “no circle” property of networks is found. Basic fuzzy tools and the related algorithms are also discussed. Some examples are analyzed to demonstrate the theory.     

用LDA Boosting算法进行客户流失预测

本文提出一种LDA boost(Linear Discriminant Analysis boost)分类方法,该算法能有效利用样本的所有特征,并且能够从高维特征空间里提取并组合优化出最具有判别能力的低维特征,使得样本类间离散度和类内离散度的比值最大,从而不会产生过度学习,大大提高算法效率。该算法有效性在某商业银行的客户流失预测过程的真实数据集中得到了验证。与其他同类算法,如人工神经网络、决策树、支持向量机等运算结果相比,该方法可以显著提高运算精度。同时,LDAboosting与其他boosting算法相比,也具有显著的优越性。     

理想簇和根理想

关于环的理想的根有两种定义,一种是所有包含I的极大理想的交,另一种是所有包含I的素理想交,本文主要研究后者定义的一些性质,以及和理想簇V(I)(所有包含I的素理想的集合)的关系.     

顾客忠诚度测评模型

简述顾客忠诚的意义 ,提出顾客心理、言语、行为忠诚的概念 ,提出顾客忠诚度的定义和测评基本原理 ,并在此基础上构建以模糊推理和模糊综合评判为主要工具的测评模型 ,且进行实证研究     

Available-To-Promise modeling for multi-plant manufacturing characterized by lack of homogeneity in the product: An illustration of a ceramic case

Lack of homogeneity in the product (LHP) appears in some production processes which incorporate raw materials that originate directly from nature and/or production processes with operations that confer heterogeneity to the characteristics of the outputs obtained, even when the inputs used are homogeneous. Poor LHP management may have a very negative impact on the customer service level and on the supply chains’ operation costs, especially when the customer needs to be served with homogeneous units of one same product. One of the key processes for suitable LHP management is the order-promising process. This work presents a mathematical programming order-promising model for make-to-stock environments with LHP. The model considers two objectives placed within a single objective by the weighted sum method. For the purpose of testing the validity of the proposed model and to evaluate the characteristics of the solutions obtained in different scenarios, numerical experiments based on realistic data from a ceramic tile company have been conducted. The results show that better results are obtained for the defined performance measures if multiple objectives are considered when promising orders than the single objective of maximizing profits. Furthermore, the superiority of the results obtained from the proposed model, if compared with current company practice, proves the model’s utility.     

Convolution Hankel Transforms on the Zemanian Spaces

In this paper we investigate the convolution Hankel transforms on the Zemanian spaces of Hankel transformable functions and distributions. The convolution Hankel transform is defined on generalized functions by using the adjoint method. Our new definition includes as special cases other known definitions of the convolution Hankel transform of distributions. Finally we establish a distributional inversion formula for the transformation under consideration involving Bessel differential operators.     

Defining continuity of real functions of real variables

Continuity of a real function of a real variable has been defined in various ways over almost 200 years. Contrary to popular belief, the definitions are not all equivalent, because their consequences for four somewhat pathological functions reveal five essentially different cases. The four defensible ones imply just two cases for continuity on an interval if that is defined by using pointwise continuity at each point. Some authors had trouble: two different textbooks each gave two arguably inconsistent definitions, three more changed their definitions in their second editions, two more claimed continuity at a point for functions not defined there, and one gave a definition implying it for a function with no limit there.     

客户关系动态优化模型与实证研究

提出了客户关系与营销活动的动态交互模型,以长期收益最大化为目标,优化企业的营销活动。模型假设客户关系可离散为几个层级状态,并设客户关系所处状态受营销活动的影响而动态的变化,服从马尔可夫决策过程。客户关系状态所处层级不可直接观测,但其与客户购买水平有概率相关关系。提出模型参数估计的最大似然估计方法。以国内某企业的客户关系管理数据为例,说明了模型变量的定义方法,通过客户交互历史数据估计模型参数,并对客户管理策略进行优化。结果表明,最优策略管理下期望提升客户价值61％～82％。     

基于两阶段DEA方法的中国商业银行效率评价

本文考虑商业银行吸收存款和发放贷款的两阶段特点,引入“顾客满意度”这一产出指标,应用两阶段DEA模型计算了我国15家商业银行在2008年和2011年的服务质量效率和盈利效率,结果显示我国商业银行的盈利效率显著高于服务质量效率。进而综合考虑服务和盈利指标,分析了各银行的综合效率,并通过计算各银行Malmquist指数研究我国商业银行2008年到2011年间的效率变化。结果表明:各商业银行从2008年到2011年的效率整体呈上升趋势,各商业银行的技术效率以及整个银行业的效率都有一定的提高。最后,应用Tobit模型分析了影响银行效率的因素,结果显示银行的贷存比、资产利润率和不良贷款率对效率影响显著。