The query time was shortened efficiently by adopting sampling method for storing historical stream data. However, important information was discarded in a large possibility because all data were treated non-discriminately with same sampling probability. In this paper, we addressed the problem by introducing a method for storing historical stream data based on degree of interest. The method used parameter Q to indicate the importance of each data, based on which a sampling algorithm had better effect on maintaining the important information. The experimental results showed that the degree of interest based method had fine performance in the query accuracy, especially for simple query on the historical stream data.