Abstract: Existing incomplete data filling algorithm are all use the same method to fill all the missing values,and did not consider the importance of each value, thus, makes all algorithms low efficiency and poor real-time. Therefore,this paper proposes a new data filling algorithm based on distinguishing the importance of attributes,it uses attribute reduction to distinguish important attributes and unimportant attributes,then,uses the improved mahalanobis-based algorithm to imputing the missing value that belong to the important attributes, and unimportant missing values according to the similarity -probabilistic method, thus,ensure that the accuracy of data,at the same time,make sure the real -time and practicality.at last,the experimental part using the Digital-home system and the UCI standard datasets to analysis the algorithm performance,verifying the superiority of the algorithm.