In this study, we analyze the portfolio selection in multiple period consumption and investment model that constructs an artificial market. In the model, there exist non-risk assets and risk assets in the market. Each agent determines redistribution of present property for maximizing the expected utility of consumption based on own judgment. Rational actions of investors who take part in the market are virtually realized on a computer by applying reinforcement learning to agents. Moreover, we propose a decision making support mechanism of portfolio selection by using reinforcement learning.