Weight determination is a key problem in studying the decision, the existing weight determination method based on rough set overcomes the shortcomings of relying too heavily on prior knowledge, but it neglects the interaction among attributes. In consideration of the limitations of equivalence relation, this paper focuses on attribute weight determination in the information system based on dominance relation, the importance of attribute determined by information quantity of the attribute set is firstly put forward. And the attribute reductions are acquired based on information quantity, which ensures the distinguish quality at the same time, and simplifies the calculation; Secondly the Shapley value method is initially chosen for the loss of information allocation, and then the attribute weights are determined. The method, taking strict km as basis, fully considers the interaction among attributes and guarantees the fair and reasonable allocation of attribute weights. The example demonstrates the specific operation steps of the Shapley value method used to determine the attribute weights. The method is easily popularized in intuitiveness and operability, provides reference for attribute weights determination in all kinds of information system, and is also a successful cross application of the game theory.