基于特征加权张量分解的标签推荐算法研究  

Tag recommendation algorithm based on feature weighting and tensor decomposition

作　　者：孙玲芳[1] 冯遵倡 

机构地区：[1]泰州学院计算机科学与技术学院,江苏泰州225300 [2]江苏科技大学计算机科学与工程学院,江苏镇江212003

出　　处：《江苏科技大学学报（自然科学版）》2015年第6期574-579,共6页Journal of Jiangsu University of Science and Technology:Natural Science Edition

基　　金：泰州市科技支撑项目(TS201515);教育部人文社科基金资助项目(10YJAZH069);江苏省"六大人才高峰"项目(2012XXRJ-013)

摘　　要：针对标签推荐系统存在极度稀疏性的问题,通过提取标注过程的关键特征并计算元组的初始权重,构建加权元组集的张量模型;然后应用高阶奇异值分解(high order singular value decomposition,HOSVD)对张量模型降维,根据处理结果作标签推荐,从而达到提高推荐效率的目的;运用Movie Lens数据集对基于特征加权张量分解的标签推荐算法进行了模拟,实验结果表明:基于特征加权张量分解的标签推荐算法比传统算法推荐效果更好。该方法能够有效改善数据稀疏性问题,提高了推荐效率.Aiming at the problem that the tag recommendation system is extremely sparse,the tensor model of weighted tuble set is constructed by extracting the key features of the tagging process and calculating the initial weights of the elements; Then,we use the high order singular value decomposition（ HOSVD） to reduce the dimension of the tensor model,So that it can improve the recommendation efficiency; The Movie Lens data set is used to simulate the tag recommendation algorithm based on feature weighting tensor decomposition. The experimental results show that the tag recommendation algorithm based on feature weighting tensor decomposition is better than the traditional algorithm. The proposed method can effectively deal with the data sparsity problem and improve the recommendation effect.

关 键 词：大众标注 标签推荐 张量分解 特征加权 高阶奇异值分解 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]