中文摘要：

为了更好地理解时效性对重复消费的影响，提出了一种包含质量和时效性的重复消费混合模型。在重复消费动力学中，用户对重复项的选取取决于先前所有相同项的时效性权重之和。在混合模型的参数求解过程中，提出了一种交互迭代的参数更新方法。由于用户的真实消费数据往往服从幂率分布，分析了幂率分布下的模型倾斜性，并给出了倾斜性的紧致下届。通过对GPlus和YouTube两个真实的数据集进行分析发现，用户的重复消费数据存在着时效性。在数据集服从幂率分布的情况下，对数据集进行拟合可以很好地对用户的重复消费进行预测。

英文摘要：

In order to better understand the effect of recency on repeat consumption, this paper proposed a hybrid model including quality and recency of items. In dynamics of repeat consumption, the selection of items for a user depended the sum of all weights from the same item. While solving the parameters of the hybrid model,this paper proposed a complementary iterative update method for parameters. Because users’ real consumption data always conformed to the power law distribution, it analyzed the tipping of the model under power law distribution, and gave a tighter lower bound for tipping. Experiments on GPlus and YouTube datasets show that,there are real recency in people5 s repeat consumption data. While the datasets conforming to the power law distribution,it can better predict people5 s repeat consumption according to fitting the datasets.