A method and device for predicting user loss

The embodiment of the invention discloses a method and device for predicting user loss, and the method mainly comprises the steps: employing the access activeness of a loss user sample in an access user sample set to a target product, screening out a key loss index influencing the user loss behavior, and enabling the loss user sample to be determined based on the definition of the loss user; Establishing a user loss prediction model by adopting a preset machine learning algorithm according to data of each access user sample in the access user sample set related to the key loss index and the sample label; And performing user loss prediction on the access user of the target product based on the user loss prediction model. Therefore, the user loss prediction accuracy of the target product canbe improved. 本说明书实施例公开了一种预测用户流失的方法和装置，主要包括：采用访问用户样本集合中流失用户样本对目标产品的访问活跃度，筛选出影响用户流失行为的关键流失指标，其中，流失用户样本是基于流失用户定义确定的；并根据访问用户样本集合中每个访问用户样本的与关键流失指标相关的数据及样本标签，采用预设机器学习算法建立用户流失预测模型；基于用户流失预测模型对目标产品的访问用户进行用户流失预测。从而，可以提升目标产品用户流失预测的准确性。