CAJ | 학술논문

针对数据稀疏性问题对于传统协同过滤推荐带来的影响,提出基于项目属性和局部优化的协同过滤推荐算法(collaborative filtering recommendation algorithm based on item attribute and local optimization,简称CUCF).算法首先改进j accard系数来优化评分的项目相似性；其次引入拉普拉斯平滑方法对基于项目属性的项目相似性进行优化；最后结合两方面的相似性结果,并且利用局部优化方法选择目标的近邻对象作为推荐群.实验结果表明,该算法减小了数据稀疏性对推荐结果的负面影响,有效地降低了预测结果的平均绝对误差 MAE.实验进一步对比了其他4种不同推荐方法,预测精度提高7.1%~15.5%,从而证明了 CUCF方法在预测准确率方面能够取得较好的效果.
침대수거희소성문제대우전통협동과려추천대래적영향,제출기우항목속성화국부우화적협동과려추천산법(collaborative filtering recommendation algorithm based on item attribute and local optimization,간칭CUCF).산법수선개진j accard계수래우화평분적항목상사성；기차인입랍보랍사평활방법대기우항목속성적항목상사성진행우화；최후결합량방면적상사성결과,병차이용국부우화방법선택목표적근린대상작위추천군.실험결과표명,해산법감소료수거희소성대추천결과적부면영향,유효지강저료예측결과적평균절대오차 MAE.실험진일보대비료기타4충불동추천방법,예측정도제고7.1%~15.5%,종이증명료 CUCF방법재예측준학솔방면능구취득교호적효과.
To overcome the impact of data sparsity on traditional collaborative filtering,we presented collaborative filtering recommendation algorithm based on item attribute and local optimization,named CUCF.We firstly used the improved j accard coefficient to optimize the similarity of item scoring.Then,we employed the Laplace smoothing method to get the similarity of item attribute.Finally,we made a linear combination of these two similarity results of items,and then used local optimization options to select neighbors as a reference for the target group. Our experimental results showed that the CUCF algorithm could reduce the negative impact of data sparsity on recommendations and effectively lower the mean absolute error of prediction consequences.Our experiments further contrast CUCF with the other four different recommendation methods,the precision of prediction was increased from 7.1% to 1 5.5%.It proved that in terms of prediction accuracy, the CUCF algorithm could achieve better results.