Missing value estimation for database of aluminophosphate (AlPO) syntheses

Database of AlPO syntheses will serve as useful guidance for the rational synthesis of microporous functional materials. But the database contains missing values about 29% of total. In this paper, to deal with the problem of missing values, four missing value estimation methods (back-propagation neural networks imputes, K-nearest neighbor imputes, singular value decomposition imputes and least square imputes) are employed on database of AlPO syntheses for the first time. The efficiency and revise of estimation methods are demonstrated by normalized root mean squared error (NRMSE) and prediction accuracy. A large number of experimental results show that the estimation methods are competent for microporous aluminophosphates and the BPimpute is recommended.

Figure optionsDownload as PowerPoint slideHighlights
► Missing values impute methods are employed to AlPO syntheses database for the first time.
► BPimpute method is proposed and exhibits the best performance.
► The missing values estimation methods can correct true data.