一种融合多种用户行为的协同过滤推荐算法

Collaborative Filtering Algorithm Integrating Multiple User Behaviors

协同过滤作为目前应用最成功的个性化推荐技术,在电子商务、社交网络等领域得到了广泛应用。然而,当此类算法应用到个性化医疗推荐领域时,由于个人医疗行为本身的复杂性和多样性,出现了推荐准确率下降的问题。针对这一问题,提出一种融合多种用户行为的协同过滤推荐算法,使用权重因子来综合衡量不同用户行为对推荐质量的影响,并引入重合依赖度的概念来修正传统的相似度度量方法。在收集的Top-md数据集上的实验结果表明,该算法能够全方位表达用户的就医偏好和意愿,有效提高个性化医疗推荐系统的推荐质量。

Collaborative filtering is one of the most successful techniques among personalized recommender systems, and is widely used in the field of e-commerce, social networks etc. Due to the complexity and diversity of the personal health behaviors, it causes low accuracy of the recommendation algorithms when applied to personalized medicine recommenda- tion. To deal with this problem, a new collaborative filtering algorithm integrating multiple user behaviors was pro- posed. The weighting factor and the overlap-dependency are introduced in classical similarity computing, and they can measure the effects of different user behaviors on the recommendation quality. Experiments on the Top-md dataset show that the new algorithm can fully express the user＇s preferences and wishes for medical treatment, and can effectively im- prove the quality of personalized medicine recommender systems.