优化稀疏数据集提高协同过滤推荐系统质量的方法

《优化稀疏数据集提高协同过滤推荐系统质量的方法》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、优化稀疏数据集提高协同过滤推荐系统质量的方法摘要:协同过滤是目前个性化推荐系统中效果较好的一种推荐技术。由于用户和项目数量的急剧增加，使得反映用户喜好信息的评分矩阵非常稀疏，严重影响了协同过滤技术的推荐质量。针对这一问题提出了综合均值优化填充方法，该方法相比较于缺省值法和众数法，考虑到了用户评分尺度问题，同时也不存在众数法中的“多众数”和“无众数”问题。在同一数据集上，通过使用传统的基于用户的协同过滤算法进行验证，表明此方法可以有效提高推荐系统的推荐质量。关键词:推荐系统;协同过滤;均值;众数;信息过载optimizationo

2、fsparsedatasetstoimprovequalityofcollaborativefilteringsystemsliuqing.peng,chenming.rui*collegeofinformationscienceandtechnology,hainanuniversity,haikouhainan570228,chinaabstract:currently,thecollaborativefilteringisoneofthemoresuccessfulandbetterpersonalizedrecomm

3、endationtechnologieswhichareappliedtothepersonalizedrecommendationsystems.asthenumberofusersanditemsincreaseddramatically,thescorematrixwhichreflectstheusers’preferenceinformationisverysparse.thesparsematrixseriouslyaffectedtherecommendedqualityofcollaborativefilterin

4、g.tosolvethisproblem,paperpresentsthecomprehensivemeanoptimalfillingmethod.themethodiscomparedtothedefaultmethodandthemodemethod,ithastwoadvantages.first,themethodtakesintoaccountuserratingscaleissues.second,themethoddoesnothavethe“multiplemode”andthe“nomode”problems.

5、onthesamedataset,byusingtraditionaluser-basedcollaborativefilteringtotesttheeffectivenessofthemethodandprovedthatthenewmethodcanimprovetherecommendedqualityofrecommendationsystems.currently,thecollaborativefilteringisoneofthesuccessfulandbetterpersonalizedrecommendati

6、ontechnologiesthathavebeenappliedtothepersonalizedrecommendationsystems.asthenumberofusersanditemsincreasedramatically,thescorematrixwhichreflectstheuserspreferenceinformationisverysparse.thesparsematrixseriouslyaffectstherecommendationqualityofcollaborativefiltering

7、.tosolvethisproblem,thispaperpresentedacomprehensivemeanoptimalfillingmethod.comparedtothedefaultmethodandthemodemethod,thismethodhastwoadvantages.first,themethodtakesaccountofuserratingscaleissues.second,themethoddoesnothavethe“multiplemode”andthe“nomode”problems.ont

8、hesamedataset,usingtraditionaluser.basedcollaborativefilteringtotesttheeffectivenessofthemethod,andtheresultsprovethatthenew