摘要
客户流失是企业面临的一个重要问题,为及时发现流失客户,降低企业损失,目前已有许多研究对客户流失问题给出解决方案,但是大部分研究中使用的是浅层学习算法,预测结果依赖于特征选择,需要在特征工程上花费大量的时间和精力。随着客户数据的快速增长,在大数据情况下,人工特征工程已不能有效地获取高质量特征。深度学习通过模拟人脑多层、逐级地抽取信息特征,能自动学习到较好的数据特征,在图像识别、语音识别等领域取得显著成果。为研究深度学习在客户流失预测方面的应用,构造了基于深度神经网络的流失预测模型,并在电信客户数据集上,与经过特征选择的Logistic回归、决策树等预测模型作对比,验证其预测准确度。实验结果表明,深度神经网络模型取得了较好的预测效果。
One of the important problem enterprise faced is customer churn. In order to find out the customer loss in time and reduce the loss of enterprises,many researchers have proposed solutions to the problem of customer churn. However,most studies use shallow learning algorithm,whose prediction results depend on feature selection and require a lot of time and energy in feature engineering. With the rapid growth of customer data,in the case of big data,artificial feature engineering has been unable to effectively obtain high-quality features. Deep learning can automatically learn better data features by simulating the human brain to extract information features in multiple layers and step by step,making remarkable achievements in the fields of image recognition and speech recognition. In order to study the application of deep learning in customer churn prediction,a churn prediction model based on deep neural network is constructed and compared with the Logistic regression,decision tree and other models after feature selection in the telecom customer data set to test the prediction accuracy. Experiment shows that deep neural network model has better prediction effect.
作者
马文斌
夏国恩
MA Wen-bin;XIA Guo-en(School of Business Administration,Guangxi University of Finance and Economics,Nanning 530003,China)
出处
《计算机技术与发展》
2019年第9期76-80,共5页
Computer Technology and Development
基金
教育部人文社会科学规划基金项目(17YJA880080)
广西跨境电商智能信息处理重点实验室培育基地(广西财经学院)专项资助项目
广西财经学院创新治理与知识产权学科群(政府治理的互联网创新发展)专项资助项目
关键词
深度学习
深度神经网络
客户流失
电信
deep learning
deep neural network
customer churn
telecommunications
