Hanabi card game is a cooperative card game. Unlike the other games, the players can't see their own cards and can only see other people's. So, it is very challenging for AI players to learn this game. In this study we simulated the Hanabi card game and trained the AI player by using the Q-learning method. However, Q-learning method will take a large amount of time if space states is numerous. Therefore, we parameterized the numbers and kinds of cards to estimate the size of the space states. Finally, we minimized the cards number and trained the AI player by using Q-learning in a short time.