CAJ | 학술논문

传统协同过滤算法中相似性度量方法存在度量不准确，不能如实反映信息的问题，导致推荐质量不高。针对这一问题，提出一种新的基于多向测度和项目属性的相似性计算方法。利用多向测度方法计算用户对项目的评分相似性，同时结合项目属性计算用户对项目属性的偏好相似度，通过加权因子得到用户间的最终相似性。实验结果表明该方法较传统方法显著提高了推荐精度。
전통협동과려산법중상사성도량방법존재도량불준학，불능여실반영신식적문제，도치추천질량불고。침대저일문제，제출일충신적기우다향측도화항목속성적상사성계산방법。이용다향측도방법계산용호대항목적평분상사성，동시결합항목속성계산용호대항목속성적편호상사도，통과가권인자득도용호간적최종상사성。실험결과표명해방법교전통방법현저제고료추천정도。
Similarity measurement algorithm in traditional collaborative filtering algorithm has the problems such as inaccurate measure and cannot truly reflect the information,thus results in low recommendation quality.To solve these problems,we propose a new similarity calculation method which is based on multi-directional measurement and project attributes.It uses multi-directional measurement to calculate the rating similarity of users on project.Meanwhile,it calculates the preference similarity of user on project attributes in combination with the project attributes,and obtains final similarity between users through weighting factor.Experimental results show that the method significantly improves the recommendation precision than the traditional algorithm.