适应情景变化的协同推荐算法

Collaborative Recommendation Algorithm Adapted to the Changing Context

协同推荐的基本思想是相似的用户往往具有相似的用户偏好,现有的个性化协同推荐算法在情景相似的前提下进行协同推荐,无法动态地适应情景变化。针对这个问题,提出一种适应情景变化的协同推荐算法。引入用户情景效用的概念,并给出了计算情景效用的有效方法,对每个具体用户,计算其对情景的效用,在经典算法的基础上,根据当前情景与历史情景的效用调整对用户的预测值,算法不用进行前置过滤,用户数据利用率更高;对每个用户计算情景效用,对用户的针对性更强;能根据情景动态调整预测值。实验表明:算法能动态适应情景变化,并提高了个性化推荐算法的推荐质量。

The basic idea of collaborative recommendation is that similar users often have similar user preferences.The existing context-ware collaborative recommendation algorithm is recommended under the premise of similar context,and cannot dynamically adapt to situation changes.Aiming at this problem,a collaborative recommendation algorithm for adapting to context changes is proposed.This paper introduces the concept of user context utility,and give an effective method to calculate the effect of the context on users.The algorithm calculates the utility of the context for each specific user.Based on the classical collaborative recommendation algorithm,the algorithm adjusts the predicted value of user based on the difference between the utility of the current and historical context.The algorithm has a higher user data utilization without pre-filtering;the context utility is calculated for each user,which is more targeted to the user;and The algorithm can dynamically adjust the predicted values based on the context.Experiments show that the algorithm can dynamically adapt to the context change and improves the recommendation quality of context-ware recommendation algorithm.