CAJ | 학술논문

数据稀疏性问题是传统的协同过滤算法主要的瓶颈之一。迁移学习利用辅助领域的用户评分信息，有效地缓解了目标领域的稀疏性问题。现有的迁移学习推荐算法中，普遍存在领域间的用户需要一致、模型平衡参数较多等限制。针对这些局限性，提出了一种用户相似度迁移的模型，利用辅助领域的用户相似度帮助目标领域用户相似度的学习。此外，通过一种用户特征子空间的距离来度量模型的平衡参数，使模型更加具有智能性。实验结果表明，该模型与其他协同过滤算法相比较能够更有效地缓解数据稀疏性问题。
수거희소성문제시전통적협동과려산법주요적병경지일。천이학습이용보조영역적용호평분신식，유효지완해료목표영역적희소성문제。현유적천이학습추천산법중，보편존재영역간적용호수요일치、모형평형삼수교다등한제。침대저사국한성，제출료일충용호상사도천이적모형，이용보조영역적용호상사도방조목표영역용호상사도적학습。차외，통과일충용호특정자공간적거리래도량모형적평형삼수，사모형경가구유지능성。실험결과표명，해모형여기타협동과려산법상비교능구경유효지완해수거희소성문제。
Data sparsity is one of the most challenges for traditional collaborative filtering algorithms . Transfer learning methods use the knowledge of from auxiliary domain so that it can effectively alleviate the data sparsity problem in target domain . However , there exist some restricts for the most existing transfer learning methods , such as users needing consistent between domains , more balanced parameters in the model etc . In response to these limitations , we propose a model of user similarity transfer , which makes use of the user similarity from auxiliary domain to help learning user similarity in target domain . In addition , we make the model more intelligent by measuring the balanced parameters in the model with distance of user feature subspace . The experimental results show that our model can more effectively alleviate the data sparsity problem compared to other collaborative filtering algorithms .