The invention provides an air conditioner fault detection and identification method and device based on a prediction and classification mode. The method comprises the steps that S101, an operation prediction model and a fault classification model are built respectively through normal state operation data and fault operation data according to the normal state operation data and the fault operationdata of an air conditioner; S102, whether the air conditioner breaks down or not is judged through the operation prediction model, if the air conditioner breaks down, the step S103 is executed, and ifthe air conditioner does not break down, the step S101 is executed; and S103, data segment where the air conditioner breaks down is located through the operation prediction model, and a fault type corresponding to the data segment is obtained through the fault classification model. According to the method, the operation prediction model and the fault classification model are built respectively through the normal state operation data and the fault operation data of the air conditioner, the state prediction model and the classification model are creatively fused, high detection and identification accuracy are achieved, meanwhile, the continuously accumulated operation data can be fully utilized, and through continuous machine learning, fault diagnosis accuracy is improved.