A method and device for predicting a user problem based on data drive. The method comprises: when receiving a problem proposed by a user, collecting user behaviour data and pre-processing same (S1); intercepting behaviour data to be selected contributive to the problem proposed by the user from the pre-processed user behaviour data (S2); and screening the behaviour data to be selected by means of a set target behaviour data set, screening out behaviour data to be selected contained in the target behaviour data set from the behaviour data to be selected, inputting the screened behaviour data to be selected into a well-trained classifier model, and predicting a category to which the problem proposed by the user belongs (S3). The device comprises a pre-processing module, an interception module and a prediction module. The method and device can significantly improve the prediction effect.
展开▼