A novel method for the recovery of continuous missing data using multivariate variational mode decomposition and fully convolutional networks

•Multivariate variational mode decomposition is employed to reduce the data complexity of original signal.•Fully convolutional networks are applied to train the mapping relationship between the decomposed subsequence components.•A systematic evaluation pattern is proposed to evaluate the recovery results from both the time and frequency domains.•Experimental study and parameter analysis prove the superiority of the proposed method.