遗忘曲线的协同过滤推荐模型

针对资源历史评价信息的时效量化问题,研究了一种应用遗忘曲线的协同过滤推荐模型.该模型以推荐信系统的构成要素为基础,将历史信息的时效量化函数引入到未知评分资源的推荐计算中.通过多阶段时效量化方法与时间单位映射函数,揭示不同资源时效衰减速率随用户兴趣变化的规律,获取具有记忆心理学中遗忘特征的推荐评价数值.实验表明,与现有协同过滤推荐模型的推荐效果相比,模型能够合理解决时效量化问题并提供质量较高的推荐效果.

Forgetting curve-based collaborative filtering recommendation model

According to the time effect of historical information in a recommendation system,a forgetting curvebased collaborative filtering recommendation system was proposed.Based on components of the recommendation system,the model involved a time-effect quantization function into the calculation process of recommendations for unknown resources.Through a multi-procedure time-effect quantization model and the unit time mapping function,the model revealed a pattern of time-effect attenuating with customer’s interest for various resources,and obtained recommendations corresponding with forgetting characteristics of memory psychology.Experimental results demonstrate that this approach can reasonably solve time-effect quantization issues and perform better than current collaborative filtering recommendation systems.