Journal of Digital Information Management Cui-cui Sun, Chun-long Yao, Xu Li, Kejun Lee 12 5 2014 http://dline.info/fpaper/jdim/v12i5/5.pdf Criminal behaviors can reflect the characteristics of the criminals to a great extent. To predict the crime types according to characteristics of vast amounts of criminals is an important part of criminal behavior analysis. In order to get high classification accuracy, three typical classification algorithms, including C4.5 algorithm, Naive Bayesian algorithm and K nearest neighbor (KNN) algorithm, are compared using several popular missing data filling algorithms respectively based on a real crime dataset with lots of missing data. The experimental results show that higher classification accuracy can be obtained by combining KNN classification algorithm and GBWKNN missing data filling algorithm which is based on grey relational analysis (GRA) theory.