结合灰色预测的动态概率矩阵分解（英文）

发布时间：2019-06-10 07:19

【摘要】：推荐系统的目标是找出符合用户喜好的物品,但是用户的喜好和物品的特征是动态变化的,这种变化会影响推荐系统的准确性.很多推荐系统只是简单的使用概率矩阵分解模型,缺乏对这个问题的有效解决.本文利用灰色系统理论中的灰色预测模型对用户和物品的动态性建模,继而提出了一个基于概率矩阵分解和灰色预测模型的动态推荐系统.首先,利用概率矩阵分解模型生成各个连续时间窗中用户和物品的隐式向量.接着,利用灰色预测模型得到未来时间窗中用户和物品的隐式向量,继而进行推荐.实验结果说明本文的算法能够有效地对用户和商品的动态性进行建模,且优于一些现存的最好的算法.
[Abstract]:The goal of the recommendation system is to find out the items that accord with the user's preferences, but the user's preferences and the characteristics of the items are dynamic, which will affect the accuracy of the recommendation system. Many recommendation systems simply use probability matrix decomposition model and lack of effective solution to this problem. In this paper, the grey prediction model in grey system theory is used to model the dynamics of users and objects, and then a dynamic recommendation system based on probability matrix decomposition and grey prediction model is proposed. Firstly, the probability matrix decomposition model is used to generate the implicit vectors of users and objects in each continuous time window. Then, the implicit vectors of users and objects in the future time window are obtained by using the grey prediction model, and then the recommendation is made. The experimental results show that the proposed algorithm can effectively model the dynamics of users and goods, and is superior to some of the existing best algorithms.
【作者单位】：中山大学数据科学与计算机学院;广东省信息安全技术重点实验室;
【基金】：Supported by National Natural Science Foundation(61502543,61573387) Guangzhou Program(201508010032) Guangdong Natural Science Funds for Distinguished Young Scholar(2016A030306014) Ph.D.Start-up Fund of Natural Science Foundation of Guangdong Province,China(2014A030310180) Fundamental Research Funds for Central Universities(16lgzd15)
【分类号】：TP391.3

【相似文献】