Incomplete data fill method in the application of pattern recognition

At present, achievements about pattern recognition has been quite rich, but most are based on the premise of complete information systems. But in the actual engineering applications, the limited factors such as measurement error and data acquisition method, people often get incomplete data. Although dealing with incomplete data has attracted wide attention and development, but as far as I know, reports on incomplete data imputation applying in pattern recognition is very rare. Aiming at incomplete information systems, This paper respectively using average imputation and regression imputation method to deal with incomplete iris data set(Part of the attribute value in the standard iris data set are abandoned artificially), and the complete data after being filled is applied to pattern recognition, by comparing recognition results with the recognition result of standard data set, the applicability of average imputation and regression imputation in the incomplete data pattern recognition is discussed.