Now a days probabilistic approach can be a convenient solution for solving many of the problems especially the problems with high time complexity like as knapsack problem. In this paper I present a new quantum computation approach based on learning strategy and greedy approach for solving the zero and one knapsack problem with polynomial time complexity about O (n) and as a result of evaluation of my presented approach I show that the probability of success of the given approach is low based on disentangled quantum registers and without learning the past observations. Nevertheless I show that how learning approach can significantly increases the probability of success of the given quantum greedy method. Also as an another result of evaluation of the presented approach I show that if the standard deviation of the items increases, the probability of success of the given new learning approach increases too.