CAJ | 학술논문

协同过滤推荐算法是目前应用最广泛个性化推荐技术，其中用户相似度的计算方法是影响推荐算法质量的关键因素。针对传统协同过滤算法中稀疏评分数据造成的用户相似度计算不准确问题，提出一种基于用户兴趣模型的协同过滤推荐算法。该算法使用分布估计算法建立用户兴趣模型，并使用用户兴趣模型计算用户间相似度。实验表明，该算法的准确性受数据稀疏性影响较小，同时在收敛速度和推荐准确性方面有明显提高。
협동과려추천산법시목전응용최엄범개성화추천기술，기중용호상사도적계산방법시영향추천산법질량적관건인소。침대전통협동과려산법중희소평분수거조성적용호상사도계산불준학문제，제출일충기우용호흥취모형적협동과려추천산법。해산법사용분포고계산법건립용호흥취모형，병사용용호흥취모형계산용호간상사도。실험표명，해산법적준학성수수거희소성영향교소，동시재수렴속도화추천준학성방면유명현제고。
Collaborative filtering recommendation algorithm is the most prevailing individualised recommendation technology.The users similarity calculation method in it is the key factor to affect the quality of the algorithm.Aiming at the issue of inaccuracy of users similarity calculation caused by the sparse rating data in classic collaborative filtering, we propose a user interest model-based collaborative filtering recommendation algorithm.It builds user interest model by using the estimation of distribution algorithms ( EDAs ) , and calculates the similaritybetween users with the model.Experiment demonstrates that the proposed approach is less affected by data sparsity on its accuracy, and meanwhile remarkably improves it convergence speed and recommendation accuracy.