This paper proposes a numerical method based on reinforcement learning that can value of option. In the celebrated option pricing model, Black Scholes, it assumes continuous hedging and zero transaction costs during hedging. This paper modifies these assumption as follows to make these assumptions as real as possible. 1) Non-zero transaction costs, 2) Discretized interval of re-hedging of an option replicating portfolio. With new methodology, this paper executed numerical experiments and examine the effects of the modified assumptions. Also, this research compares the results of the experiment with original researches of option pricing, such as Black Scholes Model and Leland Model.