基于兴趣感知和时间因子的个性化菜品推荐 *

目前推荐系统已广泛应用在各种电子商务网站上，但针对菜品的个性化推荐很少。针对菜品推荐中存在别名多、用户菜品矩阵稀疏以及新用户冷启动等难题，对基于用户的协同过滤算法进行改进，设计一种融合专家选择和在线推荐的菜品推荐系统。专家选择通过对菜品进行种类层次划分为用户兴趣建模做准备，在线推荐通过兴趣感知选择算法选择餐厅中的专家用户和候选菜品，从而实现对用户菜品的推荐。最后通过在候选菜品选择时引入时间敏感因子和协同过滤中引入时间遗忘因子，改进兴趣感知算法和菜品偏好预测效果。实验结果表明，所设计算法较传统算法在准确性和推荐效率有明显改进,并得出了针对菜品推荐时引入时间因子有利提高推荐准确性的结论。

Preference-aware and time factor based personalized dish recommendation

Currently, recommendation systems were widely used in a variety of e-commerce sites, but few were for personalized dish recommendations. Focusing on dish recommendation problems of multiple aliases, sparse user-dish matrix and cold start of new users, this paper improved the traditional user-based collaborative filtering algorithm and designed a dish recommendation system that combined expert selection and online recommendation. The expert selection made preparations for user interest model by classifying dish categories; online recommendation selected expert users and candidate dishes at restaurant by using preference-aware selection algorithm, which realized personalized dish recommendations for users. Consequently, by introducing the time-dependent factor in dish selection and the time-forgetting factor in the collaborative filtering, it improved preference-aware selection algorithm and dish preference prediction performance. The experimental results showed that the proposed algorithm improved significantly than the traditional one in terms of accuracy and recommendation efficiency, and it concludes that the introduction of time factor is beneficial to improve the accuracy of dish recommendation.