Data mining rely on large amount of data to make learning model and the quality of data is very important. One of the important problem under data quality is the presence of missing values. Missing values can occur in both at the time of training and at the time of testing. There are many methods proposed to deal with missing values in training data. Many of them resort to imputation techniques. However, Very few methods are there to deal with the missing values at testing/prediction time. In this paper, we discuss and summarize various strategies to deal with this problem both at training and testing time. Also, we have discussed the compatibility between various methods at training and testing to achieve better results.