To improve the similarity measurement between users, a similarity measurement approach incorporating clusters of intrinsic user groups（ SMCUG） is proposed considering the social information of users. The approach co...To improve the similarity measurement between users, a similarity measurement approach incorporating clusters of intrinsic user groups（ SMCUG） is proposed considering the social information of users. The approach constructs the taxonomy trees for each categorical attribute of users. Based on the taxonomy trees, the distance between numerical and categorical attributes is computed in a unified framework via a proper weight. Then, using the proposed distance method, the nave k-means cluster method is modified to compute the intrinsic user groups. Finally, the user group information is incorporated to improve the performance of traditional similarity measurement. A series of experiments are performed on a real world dataset, M ovie Lens. Results demonstrate that the proposed approach considerably outperforms the traditional approaches in the prediction accuracy in collaborative filtering.展开更多
文摘信息资源在分发共享过程中存在带宽拥塞、内容冗余等问题,播存网络借助"一点对无限点"的物理广播分发共享信息资源,对解决此类问题有独特优势.播存网络采用统一内容标签(uniform content label,UCL)适配用户兴趣和推荐信息资源,用户如何高效地获得自己感兴趣的UCL是播存网络中的关键问题.针对该问题,提出一种播存网络环境下的UCL协同过滤推荐方法(unifying collaborative filtering with popularity and timing,UCF-PT).首先,通过设定一对相似度阈值来计算用户与UCL数据的稀疏情况,根据稀疏情况决定二者对UCL评分的影响权值,并基于二者权值预测用户对UCL的评分,生成推荐结果集.其次,依据UCL热度调整推荐结果集的UCL顺序,从而使热门UCL更容易推荐给用户;最后提出UCL价值衰减函数,保证较新的UCL具备较高的推荐优先级.实验结果表明:与传统推荐方法相比,该方法不仅具有良好的推荐精度,还可保证所推荐UCL的热度与时效性,更适用于在播存网络环境下推荐UCL.
基金The National High Technology Research and Development Program of China(863 Program)(No.2013AA013503)the National Natural Science Foundation of China(No.61472080+3 种基金6137020661300200)the Consulting Project of Chinese Academy of Engineering(No.2015-XY-04)the Foundation of Collaborative Innovation Center of Novel Software Technology and Industrialization
文摘To improve the similarity measurement between users, a similarity measurement approach incorporating clusters of intrinsic user groups（ SMCUG） is proposed considering the social information of users. The approach constructs the taxonomy trees for each categorical attribute of users. Based on the taxonomy trees, the distance between numerical and categorical attributes is computed in a unified framework via a proper weight. Then, using the proposed distance method, the nave k-means cluster method is modified to compute the intrinsic user groups. Finally, the user group information is incorporated to improve the performance of traditional similarity measurement. A series of experiments are performed on a real world dataset, M ovie Lens. Results demonstrate that the proposed approach considerably outperforms the traditional approaches in the prediction accuracy in collaborative filtering.
