【摘要】 针对基于内容的推荐算法在进行新闻推荐时,存在时效性不足,缺乏用户兴趣多样化的问题,提出一种基于融合偏好的新闻推荐算法。首先利用基于内容的推荐算法,考虑时间对用户兴趣的影响,对用户兴趣权值的计算进行改进,发现用户的自身偏好。其次考虑时间效用和新闻热度与用户兴趣的相关性,提出一种新的混合相似度计算方式,寻找用户的最近邻,发现用户的潜在偏好。最后将用户的自身偏好与潜在偏好进行融合,产生推荐列表。实验结果表明,相比于传统推荐算法,改进后算法的平均误差绝对值最高下降了9.13%,推荐准确率最大提高了9.80%,召回率最大提高了10.12%,多样性最大提高了3.44%。改进后的推荐算法有效地提高了推荐质量,改善了缺乏多样性的问题。更多还原

【Abstract】 To address the problem that content-based recommendation algorithm is short of timeliness and lacks diversified user’s interests,a news recommendation algorithm based on mixed preferences is proposed.Firstly,the content-based recommendation algorithm is used to improve the computation of user interest weights by considering the effect of time on user’s interest and to discover users’ own preferences. Secondly,considering the relevance of time utility and news popularity to the users’ interests,a new hybrid similarity calculation method is proposed. This method can find the nearest neighbor and discover the user’s potential preferences. Finally,the user’s own preferences are coupled with the potential preferences to generate a recommended list. The experimental results show that compared to the traditional recommendation algorithm,the new algorithm exhibits a maximum decrease of 9.13% in MAE,a maximum increase of 9.80% in recommendation accuracy,a maximum increase of 10.12% in recall,and a maximum increase of 3.44% in diversity. The new recommendation algorithm effectively improves the quality of recommendation with improved diversity.更多还原