互联网推荐系统中的协同过滤算法研究-计算机软件与理论专业论文.docxVIP

回 回 浙江大学硕士学位论文 摘要 最近邻协同过滤算法结合效果最好，与传统最近邻协问过滤算法相比准确率布所 提高。 关键词z 互联网，推荐系统，协问过滤，时间权重 浙证大学研士学位论文 Abstract Abstract Usually，users have to fàce a 1IDUI1饱in of information on their own in many internet applications. No matter familiar or not wíth this mass informatio n， people hope to get useful and personalization in岛rmation from that as quickly 部 they can Recommender sy伽m has been brought fòrwardωhandle this situation Based the h岳阳rical information of users，it can make re∞mmendation which b四t meet their d创e and filter the in岛rmation that 四le部协 them One of the Ill)st typical intemet applications of reωmm盼e3剖阳Il时Cd衍er s归tωem 臼1sE pμl町ay arole ofs栩a阳le盼Sωgiveadv叽icωe制a州I 时 阳tle协白 pωplemake be创tte即r c加l1盼0比1ce. Reωmmendation algorithm is the ll10st important part of recommender system， which predicts \阴阳， rating for iterns or reωmmends od 加m to user according 协 use邸· profile ，historical information and also the profile of items. AlIDng all the r削mmendation al rit胁， colla协rative filtering is the most pop毗r one. Although there are a 1例。f su::cessful applications of reωmmender system in intem叽 00明ver， with the developrrent of web techoology and the changing of user demar时， ∞llaborative filtering is fu.cing these t阴阳ues: l)Inco甲or脚 wíth the profile of user anditemωenhance recommendation precision 啊le profile of user usually refers 协 human dellDgraphy infòrmatìl吼叫lile the profile of item includes shape and visit tir? and SO on 2) Users pre也rence is always cbangingωtime shifting. If the al rithm fails 阳岛llow the step of use邸， it ∞uld not make ∞rrect reωmmendations. 80 how to capt? the changingprererence needωbe 剑叫时， In orderωhandle the problems mentioned above，附propose a time weight based collaborative filtering al rithm This algorithm introdu::es time weightωimprovethe recommendation ∞mpute process based tr时itional neighborhood-based collaborative filtering algorithm ηle time weig}业负