The existing recommender systems often use the nearest neighbors' preference behavior to predict current users' preference, and their recommendation accuracy are influenced by the lack of consideration that users' preference would change over time. To solve this problem, a cooperative preference prediction method based on time attenuation and preference fluctuation was proposed. First, attenuation increment and attenuation speed were obtained based on time and historical preference, and the attenuation function was generated by attenuation increment and attenuation speed to modify users' historical preference behavior. Then the distribution of historical preference was used to compute the preference fluctuation range. Finally, the recommender list was generated for user by applying the attenuation function and preference fluctuation range into the acquisition of nearest neighbors and the preference acquisition process. The experimental results on real data set show that, compared with the Collaborative Filtering based on Rating Distribution (RDCF) and Optimizing Top- N Collaborative Filtering (OTCF), the average Mean Absolute Error (MAE) of the proposed method is decreased by about 6.42% and 7.73% respectively. It also shows that the proposed method can achieve higher recommendation accuracy and better recommendation quality.