In order to solve the problem that the existing methods had low accuracy when recovering serious missing multivariable time series, a missing data recovery method based on fused prior information was proposed. This method was derived from the probability matrix factorization algorithm, making full use of the correlation between multi-sensor data and the correlation between time series data. Firstly, the proposed method exploited the latent correlation between multi-sources, and constructed the approximate representation of sensor latent factor feature matrix. Then, the time series of same sensor was analyzed, and the approximate representation of time series latent factor was constructed based on similarity between time series data. Finally, the two approximate representations were unified in the framework of probability matrix factorization algorithm, the latent feature representation and PMF algorithm parameters were obtained by learning. Simulation results show that the algorithm can recover data effectively.
