The invention relates to the technical field of electric power devices and specifically relates to a user behavior analysis-based electric quantity prediction method. Data of a remote data warehouse is collected and stored in a local database, operation slowdown caused by direct access can be prevented, an attribute definition is given to user data in the remote data warehouse, attribute classification of collected data can be facilitated, corresponding user attribute information can be listed according to a mere attribute type during system designing operation, a plurality of electric quantity prediction algorithm packages are stored in the local database, conditions of all the algorithm packages can be compared conveniently, the collected data can be preprocessed and pre-analyzed via a processing controller, electric quantities of users can be predicted via the electric quantity prediction algorithm packages, real electric quantity values in the remote data warehouse can be obtained,prediction accuracy can be calculated based on the real electric quantity values and predicted electric quantity values, the prediction accuracy and the predicted values are sent to a visualized interaction interface, a suitable electric quantity prediction algorithm can further be obtained, and electric quantity prediction accuracy can be improved.