This project is about Cancer AML/ALL research. There are two phases. In the first step, we read different literatures using famous gene expression data 'Golub data' to explore machine learning classifier that has high accurate rate in predicting whether a patient has AML or ALL. We chose several papers, and implemented the classifier by the step given in papers, and calculated type I and type II error rate for each classifier using different feature selection method. In the second step, we found other datasets which also about AML/ALL research. And use Cross-Study validation on each classifier and get error rate for each classifier.