Research on missing value estimation in data mining

Data missing in data pretreatment has a serious effect on the accuracy of subsequent analysis results in data mining. In this paper, the cause of data missing and the corresponding effect on data mining were discussed. Then the missing pattern and the application limitation of traditional missing value estimation were analyzed as well. Finally, the estimation algorithm of missing value in MAR (missing at random) pattern was mainly studied and applied in fault classification based on wavelet neural network. The experiment results show the efficiency of the algorithm.