综合个人兴趣和平均偏好的矩阵分解推荐算法

Matrix Factorization Recommendation Algorithm Combining Personal Interests and Average Preferences

传统的兴趣推荐模型提升用户推荐精度的同时增加了计算耗时,没有考虑商品本身的平均偏好特性,导致推荐系统的表现不够理想。为解决上述问题,文中提出采用预分解方法快速计算得出用户的兴趣向量,可以加快模型运算速度,同时提升预测结果精度;并且论文进一步提出在模型中引入物品的平均偏好来改进top-N推荐算法,可以提升推荐结果的质量和效果。运用Netflix以及MovieLens数据集验证,实验结果表明,改进后的模型在推荐结果的质量方面优于原有模型,可以有效提高推荐效率,改善推荐准确度。

The traditional interest recommendation model improves the accuracy of user recommendation,while it also increases the calculation time,and does not take advantage of the average preference of goods. In order to solve these problems above,this paper proposes to use the pre-factorization method to quickly calculate the user’s interest vectors,which can speed up the model operation speed and improve the accuracy of prediction results. In addition,the paper further proposes that the top-N recommendation algorithm can be improved by introducing the comment preference of items into the model,which can improve the quality and effect of recommendation results. This paper uses data sets from Netflix and MovieLens to test the new model. The final experimental results show that the new model can improve the accuracy and efficiency of recommendation results effectively.