How Trade Partners Make Their Decision in Cyberspace: A Research Based on Stochastic Games

In E-commerce world, participantspsila action selection depends on other participantspsila choice. We make use of a popular method in computer branch to analyze behaviors of people in Internet and bring forward a different view to the issue. Conflict game is one kind of situations which happens frequently. Even if allowing interaction or negotiation, participants are hard to reach agreement under this kind of conflicts, which descends system performance. We adopt a rational but conservative action selection method, namely, minimizing regret function in worst case. By this method lost incurred possibly in future is lowest under this very policy, and Nash equilibrium mixed policy is obtained when without information about other agents. Based on regret reinforcement learning model and algorithm for conflict game under multi-agent complex environment are put forward. Additionally, introduce Nash-q to analyze the general-sum game and adapt itself to no-zero-sum situation in E-commerce.