基于混合数据挖掘模型预测用户流失

用户流失预测问题广泛应用在银行、金融、电信等多种领域。对用户行为进行有效的预测和分析有助于企业的竞争和了解瞬息万变的市场规律。采用3种混合的数据挖掘模型对用户流失问题进行了研究,以形成一个准确高效的用户流失预测模型。这3种模型应用于数据挖掘的两个阶段:聚类阶段和预测分析阶段。在第1阶段中,对用户的数据进行过滤。第2阶段对用户行为进行预测。第1个模型采用了二分k-means算法进行数据过滤和多层感知人工神经网络(MLP-ANN)相结合进行预测。第2个模型采用层次化聚类与MLP–ANN相结合进行预测。第3个模型使用自组织映射(Self-Organizing Maps)与MLP-ANN进行预测。这3种模型预测分析基于真实数据,用户流失率采用3种模型混合计算的方式得出结果并同真实值进行比较。分析结果表明采用多模型的混合数据挖掘模型的数据准确度优于普通的单一模型。

Anticipation of Customer Churn Based on Hybrid Date Mining Models

Customer churn prediction is widely used in a variety of fields including banking, finance and telecommunications. User behavior contributes effectively to predict and analyzes the law and understand the competitive fast-changing market. In this paper,using three hybrid data mining models,churn conducted a study in order to form an accurate and efficient customer churn prediction model. Two stages of the three models applied to data mining:clustering stage and predictive analysis phase. In the first stage,the customer's data is filtered. The second phase is customer behavior prediction. The first model is applied for the bisecting k-means algorithm for data filtering and Multilayer Perceptron artificial neural network(MLP-ANN)combined forecast. The second model uses hierarchical clustering and MLP-ANN combined forecast. The third model is the use of self-organizing map (Self-Organizing Maps)and MLP-ANN to predict. All three models predict based on real data, customer churn by the method of three hybrid models calculated and compared with the outcome of the true value. The results showed that the use of comparison data accuracy multiple models mixed data mining model is superior to conventional single model.