There are a lot of serious data quality problems in real world datasets: incomplete, redundant, inconsistent and noisy. Missing data is a common issue in data mining and knowledge discovery. It is well accepted that many real-life datasets are full of missing data. However, data mining algorithms always handle missing data in very simple way. Missing data handling has become an acute issue. Inappropriate treatment will reduce the performance of data mining algorithms. For classification algorithm, its classification accuracy depends vitally on the quality of the training data.