Existing algorithms are of low efficiency and effectiveness in imputing missing data. Aiming at this problem, the paper proposes a missing value imputation algorithm based on the CFS clustering and improved auto-encoder model. To cluster the incomplete data set, it improves the CFS clustering algorithm by introducing the partial distance strategy that is used to measure the distance between two objects with missing values. It uses the improved CFS algorithm to cluster the data set. The improved auto-encoder is used to estimate the missing values according to the clustering result. Experiments demonstrate that this proposed algorithm can impute the missing values effectively.