Materials and Methods: In this study, patients data were collected from Flow Cytometry tests of the Iran Blood Transfusion Organization (IBTO) have been used. 304 different diagnosed samples in 8 classes of acute leukemia were investigated. Samples were initially in numerical format. In the next stage, we transformed them into Boolean format according to the defined threshold. Then, weights were assigned to these samples based on cooperative game theory and Shapley value. In this regard, different samples of acute leukemia were separated and classified (Learning Phase). In the diagnosis phase, using similarity measures, the similarities between new under study and the training samples were assessed and the type of under study leukemia were detected (Diagnosis phase). Results: The accuracy rate of the classification method based on the cooperative game theory for leukemia was 96.3% which indicates that the proposed method has a considerable precision rate to classify the different kind of classes. In order to find the validity and efficiency of the proposed method, the results were compared with neural network, which is one of the useful learning algorithms. The accuracy rate of the classification method based on Radial Basis Function method (RBF) was 91.80%. Conclusion: Considering the data, the proposed method gave very hopeful results for acute leukemia classification. In this regard, it can assist hematologists and physicians in reasonable and accurate diagnosis of the kind of leukemia, to make more suitable decisions.