Abstract: Analyzing historical user behavior data and establishing user preference model by some certain method is the critical subject with great attention. This paper considers the time series characteristics of data generation and the influence of the variables with temporal characteristics on user behavior models. Based on the forgetting curve in psychology, this paper starts from user behavior data and gives the representation of user preferences. Thus, a forgetting curve model can be established for each preference and user preferences can be represented by real time manner. Aiming at massive user behavior data, this paper proposes the MapReduce-based algorithms for the incremental    update of model parameters and the computation of dynamic user preferences. Thus, inherently dynamic user preferences can be reflected by the constructed user behavior model. The experimental results conducted on real data show that the proposed model and algorithms are efficient, correct and applicable.

Key words: dynamic user behavior model, user preference, forgetting curve, incremental update, MapReduce

摘要： 分析用户行为的历史数据，使用特定方法建立用户的偏好模型，是目前研究的热点和关键。考虑了数据产生的时序特征，以及具有时间特征的变量在用户行为模型中的影响，以心理学中的记忆曲线模型为依据，从用户的行为数据出发，给出了用户偏好的表示，并为用户的每个偏好建立一个记忆曲线模型，实时地表示用户的偏好。针对海量的用户行为数据，提出了基于MapReduce的模型参数增量更新算法和动态用户偏好计算方法，从而使得模型能反映动态变化的用户偏好。建立在真实数据上的实验结果表明，提出的模型和算法具有高效性、正确性和可用性。

关键词: 动态用户行为模型, 用户偏好, 记忆曲线, 增量更新, MapReduce