Aiming at the problem that the risk warning of optical communication network equipment cannot be analyzed manually, a failure prediction algorithm of equipment based on user profile and deep learning algorithm is proposed. Based on data collection and data enhancement, the device health profile is constructed, and the label sequence associated with the equipment failure is extracted from the original data containing dirty data and inconsistent format, and the sequence data is fed into the deep learning model to obtain highly accurate failure prediction results. The simulation results show that this algorithm can construct the equipment health profile based on the original equipment data and train the deep learning model in real time, and obtain the accuracy rate of equipment failure prediction close to 100%. Compared with the algorithm of full label sequence and the algorithm without data enhancement, the proposed algorithm improves the accuracy rate of failure prediction by 7.8% and 3.3%.