Many e-commerce sites employ collaborative filtering techniques to provide recommendations to customers based on the preferences of similar users. However, as the number of customers and products increases, the prediction accuracy of collaborative filtering algorithms declines because of sparse ratings. In addition, the traditional recommendation approaches just consider the item's attributes and the preference similarities between users; however, they are not concerned that users' preferences may be developed as their familiarity with or experiences during choice or preference elicitation grows. In this work, we propose an anchor-based hybrid filtering approach to capture the user's preferences of movie genres interactively and then achieve precise recommendations. To conduct this experiment, we recruited 30 users with different types of preference stabilities for movie genres. The experimental results show that the proposed anchor-based hybrid filtering approach can effectively filter out the users' undesired movie genres, especially for the user who has unstable movie genre preferences. The results suggest that the factor of the stability of users' preferences can be considered for developing effective recommendation strategies.