Due to the development of computer and Internet technology, cooperative learning on the Web has became more and more popular. This thesis proposes the improved dynamic grouping strategy for comcept graph based cooperative learning to enhance the learners’ learning efficiency by combining the learning achievement evaluation and the features of thinking styles. The grouping strategy proposed in this thesis firstly divides the learning activities into several phases. Then, the SPRT and partial learning evaluation are used to examine the learning achievement and calculate the complementary degree of the conceptual graph in each phase. The dynamic grouping strategy is then carried out by means of combing the thinking styles obtained from the relevant questinnaire and group complementary score. The results of experiment reveal that the proposed grouping strategy can efficiently improve the learning achievement.