Monte Carlo tree search is used in AlphaGo Zero which is a strong artificial intelligence player for Go, and therefore it has attracted much attention. It can be used for searching for the optimum of combinatorial optimization problems, and it is promising for finding the optimum or a near-optimum. This paper proposes a Monte Carlo tree search method for the knapsack problem which is one of the typical combinatorial optimization problems. In the proposed method, a new candidate solution is generated by using superior ones found so far in the procedure called the simulation. Its performance is evaluated through conducting numerical experiments.