基于不平衡数据集的客户流失预测研究*

目前客户流失预测面临的主要问题之一就是类不平衡性(class imbalance)。针对这个问题,首先应用欠抽样法(undersampling)处理客户流失数据降低不平衡性,再应用C4.5D、C4.5N、RIPPER、NaiveBayes和RandomForest机器学习方法对客户流失进行预测。实验结果表明,欠抽样法是在牺牲负类样本预测精度的前提下,提高正类预测精度,于是采用重复抽样法(resampling)来弥补欠抽样法的缺陷,减少负类样本中含有大量有用信息的丢失,实验结果证明了这种方法的正确性和有效性     

利用ClementineC5.0模型预测CDMA客户流失

该文针对目前电信行业中一个日益严峻的问题:客户离网进行研究,通过收集客户的基本数据、消费数据和缴费行为等数据,建立客户流失预测模型,进行客户流失分析及预测.通过对大量相关技术和统计方法的研究,最终确定了clementine的C5.0模型作为电信客户流失的预测模型.此模型对客户流失预测有较高的准确性,为电信经营分析系统作...     

基于贝叶斯网络的电信客户流失预测分析

电信客户流失分析常用的数据挖掘方法有自动聚类、决策树和人工神经网络,它们是采用数据本身来训练模型的,没有利用先验知识。电信客户流失是由客户心理、服务质量和对手竞争等诸多复杂的因素造成的,利用这些已有的先验知识,可以提高预测的精度。该文根据先验知识选取分析变量,采集样本数据,通过贝叶斯网络的结构学习和参数学习,建立客户流失模型并进行客户流失趋势预测,取得了比标准数据集更准确的结果,该结果和决策树方法的预测结果相比还具有较大的优势,说明贝叶斯网络是分析客户流失等不确定性问题的有效工具。     

基于代价敏感SVM的电信客户流失预测研究*

针对客户流失数据集的非平衡性问题和错分代价的差异性问题,将代价敏感学习应用于Veropoulos提出的采用不同惩罚系数的支持向量机,建立客户流失预测模型,对实际的电信客户流失数据进行验证。通过与传统SVM、C4.5和ANN对比研究,结果显示此方法在精确度、命中率、覆盖率和提升度均有所改善,表明此方法有效地解决了数据集的非平衡性和错分代价问题,是进行客户流失预测的有效方法。     

改进支持向量机在电信客户流失预测的应用

电信流失客户数据精确预测是挽留客户的有效手段.电信业的管理中对收费、投诉、业务受理等问题,显然是一种典型的非平衡样本,传统用标准的支持向量机没有考虑样本分布不平衡问题,虽然在样本数据平衡前提下具有较好的预测精度,但对于不平衡电信客户数据,预测精度大大下降.为提高预测精度,针对支持向量机处理不平衡样本时的缺陷,提出了基于代价敏感学习的支持向量机模型.模型利用代价敏感学习对不平衡样本集分别采用不同惩罚系数,然后建立电信客户流失预测模型,最后对实际电信客户流失数据进行测试.通过与标准支持向量机、神经网络对比,结果表示模型提高了预测精度,有效地解决了数据集非平衡性问题,是一种有效的电信客户流失预测方法.     

基于并行C4.5的铁路零散白货客户流失预测研究

为了提高铁路零散白货客户流失预测的准确性和高效性,根据铁路零散白货客户的流失特征,提出了基于CDL模型的客户流失识别方法,在此基础上,针对数据量大的问题,提出了基于Hadoop并行框架的C4.5决策树客户流失预测模型。通过仿真实验,证明该模型具有较好的准确性和预测能力,并且随着样本数量的增加,Hadoop并行框架的效率得到了明显的提升,且不影响客户流失预测模型的准确性和预测能力。     

基于社交网络分析和XGBoost算法的互联网客户流失预测研究

针对互联网行业的客户流失预测问题,提出了一种社交网络分析和机器学习相结合的客户流失预测方法。考虑到社交活动对用户流失的影响,首先采用社交网络分析方法从用户社交数据中提取特征,然后使用XGBoost(Extreme Gradient Boosting)算法来进行客户流失预测,最后将该方法与其他机器学习算法(Logistic回归、支持向量机和随机森林)进行比较。实验结果表明,所提出的社交网络分析和XGBoost相结合的客户流失预测方法优于传统方法。     

电子商务客户流失的建模与预测研究

研究电子商务客户流失预测问题,电子商务客户流失具有非线性、时变等特点,用单一预测模型难以对电子商务客户流失变化规律进行全面、准确预测,导致预测正确率低。为了提高电子商务客户流失预测正确率,提出一种组合的电子商务客户流失预测模型。组合预测模型首先采用遗传算法对影响客户流失因子进行筛选,提取对预测结果影响重要的因子,然后分别采用支持向量机和神经网络对其进行预测,最后采用支持向量机对两种预测结果进行融合,得到组合模型的电子商务客户流失预测结果。仿真结果表明,组合模型提高了电子商务客户流失预测正确率,解决了单一预测模型的缺陷,将为电子商务客户流失研究提供一种新预测思路。     

基于信息融合的电信客户流失预测研究

针对数据挖掘方法在电信客户流失预测中的局限性,提出将信息融合与数据挖掘相结合,分别从数据层、特征层、决策层构建客户流失预测模型。确定客户流失预测指标;根据客户样本在特征空间分布的差异性对客户进行划分,得到不同特征的客户群;不同客户群采用不同算法构建客户流失预测模型,再通过人工蚁群算法求得模型融合权重,将各模型的预测结果加权得到预测最终结果。实验结果表明,基于信息融合的客户流失预测模型确实比传统模型更优。     

支持向量机在电信客户流失预测中的应用研究

客户流失分析与预测是客户关系管理的重要内容.由于电信客户的特征呈高度非线性、严重冗余和高维数,传统方法无法消除数据之间冗余和捕获非线性规律,导致预测精度较低.为了提高电信客户流失预测精度,提出一种基于主成份分析(PCA)支持向量机(SVM)的电信客户流失预测方法(PCA-SVM).首先利用主成分分析对原始数据进行特征降维,消除冗余,然后将得到的主成分作为非线性支持向量机的输入进行学习建模.对某电信公司客户流失数据进行了仿真,实验结果表明,PCA-SVM获得的命中率、覆盖率、准确率和提升系数远远高于其它预测方法.说明主成分分析结合支持向量机的数据挖掘方法具有很好的预测效果,为电信客户流失预测提供了一种新方法.     

基于GA-SVM的银行客户流失预测分析

客户流失分析与预测是客户关系管理的重要内容。根据银行实际客户流失数据中正负样本数量不平衡而且数据量大的特点,采用遗传算法对传统支持向量机进行改进,得到GA-SVM模型,并以国内某商业银行VIP客户流失预测为实例,与人工神经网络、决策树、逻辑回归和贝叶斯分类器方法进行了对比,发现该方法能获得最好的正确率、命中率、覆盖率和提升系数,是预测现有客户流失倾向的有效方法。     

基于多模型融合的流失用户预测方法

准确的用户流失预测能力有助于企业提高用户保持率、增加用户数量和增加盈利。现有的流失用户预测模型大多为单一模型或是多个模型的简单融合,没有充分发挥多模型集成的优势。借鉴了随机森林的Bootstrap Sampling的思想,提出了一种改进的Stacking集成方法,并将该方法应用到了真实数据集上进行流失用户的预测。通过验证集上的实验比较可知,提出的方法在流失用户F1值、召回率和预测准确率3项指标上均好于所有相同结构的经典Stacking集成方法;当采用恰当的集成结构时,其表现可超越基分类器上的最优表现。     

Combined rough set theory and flow network graph to predict customer churn in credit card accounts

Customer churn has become a critical issue, especially in the competitive and mature credit card industry. From an economic and risk management perspective, it is important to understand customer characteristics in order to retain customers and differentiate high-quality credit customers from bad ones. However, studies have not yet adequately introduced rules based on customer characteristics and churn forms of original data. This study uses rough set theory, a rule-based decision-making technique, to extract rules related to customer churn; then uses a flow network graph, a path-dependent approach, to infer decision rules and variables; and finally presents the relationships between rules and different kinds of churn. An empirical case of credit card customer churn is also illustrated. In this study, we collect 21,000 customer samples, equally divided into three classes: survival, voluntary churn and involuntary churn. The data from these samples includes demographic, psychographic and transactional variables for analyzing and segmenting customer characteristics. The results show that this combined model can fully predict customer churn and provide useful information for decision-makers in devising marketing strategy.     

Reconciling performance and interpretability in customer churn prediction using ensemble learning based on generalized additive models

To build a successful customer churn prediction model, a classification algorithm should be chosen that fulfills two requirements: strong classification performance and a high level of model interpretability. In recent literature, ensemble classifiers have demonstrated superior performance in a multitude of applications and data mining contests. However, due to an increased complexity they result in models that are often difficult to interpret. In this study, GAMensPlus, an ensemble classifier based upon generalized additive models (GAMs), in which both performance and interpretability are reconciled, is presented and evaluated in a context of churn prediction modeling. The recently proposed GAMens, based upon Bagging, the Random Subspace Method and semi-parametric GAMs as constituent classifiers, is extended to include two instruments for model interpretability: generalized feature importance scores, and bootstrap confidence bands for smoothing splines. In an experimental comparison on data sets of six real-life churn prediction projects, the competitive performance of the proposed algorithm over a set of well-known benchmark algorithms is demonstrated in terms of four evaluation metrics. Further, the ability of the technique to deliver valuable insight into the drivers of customer churn is illustrated in a case study on data from a European bank. Firstly, it is shown how the generalized feature importance scores allow the analyst to identify the relative importance of churn predictors in function of the criterion that is used to measure the quality of the model predictions. Secondly, the ability of GAMensPlus to identify nonlinear relationships between predictors and churn probabilities is demonstrated.     

Theoretical and experimental analysis of a two-stage system forclassification

We consider a popular approach to multicategory classification tasks: a two-stage system based on a first classifier with rejection followed by a nearest-neighbor classifier. Patterns which are not rejected by the first classifier are classified according to its output. Rejected patterns are passed to the nearest-neighbor classifier together with the top-h ranking classes returned by the first classifier. The nearest-neighbor classifier, looking at patterns in the top-h classes, classifies the rejected pattern. An editing strategy for the nearest-neighbor reference database, controlled by the first classifier, is also considered. We analyze this system. Moreover, we formally relate the response time of the system to the rejection rate of the first classifier and to the other system parameters. The error-response time trade-off is also discussed. Finally, we experimentally study two instances of the system applied to the recognition of handwritten digits. In one system, the first classifier is a fuzzy basis functions network, while in the second system it is a feed-forward neural network. Classification results as well as response times for different settings of the system parameters are reported for both systems     

GSM churn management by using fuzzy c-means clustering and adaptive neuro fuzzy inference system

Churn management is important and critical issue for Global Services of Mobile Communications (GSM) operators to develop strategies and tactics to prevent its subscribers to pass other GSM operators. First phase of churn management starts with profile creation for the subscribers. Profiling process evaluates call detail data, financial information, calls to customer service, contract details, market details and geographic and population data of a given state. In this study, input features are clustered by x-means and fuzzy c-means clustering algorithms to put the subscribers into different discrete classes. Adaptive Neuro Fuzzy Inference System (ANFIS) is executed to develop a sensitive prediction model for churn management by using these classes. First prediction step starts with parallel Neuro fuzzy classifiers. After then, FIS takes Neuro fuzzy classifiers’ outputs as input to make a decision about churners’ activities.     

基于核主成分分析特征提取的客户流失预测

将核主成分分析(KPCA)引入到客户流失预测中,提出了相应的特征提取算法。将KPCA与Logistic回归结合,设计了预测模型。通过对某电信公司客户流失预测的试验结果表明：该方法获得的命中率、覆盖率、准确率和提升系数高于原始属性集和主成分分析(PCA)特征提取法。这表明KPCA能提取客户数据的非线性特征,是研究客户流失预测问题的有效方法。     

基于联合决策树的客户流失预测模型设计

为了解决电信行业客户流失预测模型中流失者和未流失者比例偏斜问题,模型依据数据挖掘原理,以CRISP-DM（Cross-industry Standard Process for Data Mining）建模过程为框架,采用了多基决策树联合决策的思想。模型避免了训练出一棵“空”决策树,把所有客户都预测为未流失的问题。与单个分类器相比,提高了预测模型的查准率和泛化能力。     

Churn prediction and social neighbour influences for different types of user groups in virtual worlds

In virtual worlds (VWs), users have more VW games alternatives, whereas VW companies consequently suffer from high customer turnover rate and low customer loyalty. Therefore, building a churn prediction model to facilitate subsequent churn management and customer retention is important. The churn behaviours and the impact of social neighbour influences to customer churn may be different for different types of users. Accordingly, we segment users into stable, unstable, and solitary user groups according to their social contact behaviours in VWs. Novel segmentation‐based churn prediction approaches are proposed for churn prediction in VWs by building prediction models for each type of user groups and considering the effect of social neighbour influences for different user groups. The proposed approaches are evaluated by conducting experiments with a dataset collected from a VW platform. The experimental results show different churn prediction performances under different user groups. The segmentation‐based churn prediction approaches perform better than do general approaches without considering user groups. Moreover, the results also reveal that social neighbour influences have a positive impact on stable and unstable users. The proposed work contributes to investigating the social neighbour influences on churn prediction for different types of user groups in VWs.