Abstract: Concerning the limitations of data mining methods in telecom customer churn prediction, a combination of information fusion and data mining technology is proposed by establishing customer churn prediction model on data level, feature level and decision-making level. Firstly, the customer churn predictors are determined. Secondly, customers are divided into groups with different features according to their differences in the feature space distribution. Lastly, customer churn prediction results of each group are obtained by weighted the predictions of different models which are established by using different methods, and the weight of each model is obtained through artificial ant colony algorithm. The experimental results show that the performance of customer churn prediction model based on information fusion is better than the traditional model.

Key words: customer churn, data mining, information fusion, artificial ant colony algorithm

摘要： 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。

关键词: 客户流失, 数据挖掘, 信息融合, 人工蚁群算法