The invention relates to the field of arrearage power-off and power-on complaint tendency prediction, and specifically relates to an arrearage power-off and power-recovery complaint tendency prediction method based on an entropy method. The method comprises: constructing a complaint tendency prediction model based on an entropy method; and obtaining a source data width table, taking a final indexand a value thereof included in the source data width table as input values of the complaint tendency prediction model, calculating a final comprehensive score after standardization of each user according to the complaint tendency prediction model, and if the final comprehensive score of the user is higher than a first threshold value, judging that the user has the complaint tendency. Whether theuser is a sensitive user or not is judged according to an entropy method, customer complaint characteristics and change rules can be recognized, customer complaint risks are predicted, complaint pressure is effectively reduced, the service depth is effectively expanded, the working efficiency is improved, and the service level is improved.