摘要
协同过滤算法是现在个性化推荐领域流行的算法。对常见的推荐问题,协同过滤算法已有成熟的实现。单类协同过滤问题是推荐领域的一个新问题,其数据特征导致其不适用于常见的协同过滤算法。本文研究了基于加权矩阵分解的单类协同过滤算法,并对其进行基于迁移学习的改进。通过在真实数据集上的验证,证明其效果优于传统的单类协同过滤算法。
Collaborative filtering is a useful algorithm for problems of personalized recommendation. For these prob-lems, there are many mature collaborative filtering algorithms. One class collaborative filtering is a new field of per-sonalized recommendation. Because of its data characteristics, common collaborative filtering algorithms have a lot of defects in the field of one class collaborative filtering. We studied the algorithm based on weighted matrix decomposi-tion, and optimized this algorithm by transfer learning. We prove the improvement of this optimization by experiments.
出处
《数据挖掘》
2013年第1期12-17,共6页
Hans Journal of Data Mining
