推荐系统的准确度经常受到各类偏差的影响，流行度偏差是影响推荐准确度的重要因素之一。传统的纠偏方法主要基于项目属性，通过引入惩罚因子来抑制热门项目的推荐，未考虑用户兴趣和时间的影响。针对此问题，提出了基于项目流行度和用户动态兴趣的自适应纠偏方法(Adaptive Popularity and Dynamic Interest，APDI)。结合因果图从项目流行度和用户个性化两个方面综合分析影响流行度偏差的主要因素，根据项目质量、从众效应、用户兴趣对时间的敏感度不同，计算相应时间间隔内项目流行度与用户动态兴趣的综合评分，更加有效地降低流行度偏差;通过指数加权移动平均的方法，根据时间衰减程度对用户当前兴趣的影响来计算用户兴趣评分，以捕捉用户的短期兴趣偏好。在3个真实数据集上验证了所提方法的有效性，实验结果表明，APDI有效提高了推荐的准确度、召回率和归一化折损累计增益。
Debiased Recommendation Based on Item Popularity and User Dynamic Interest
The accuracy of recommendation system is often affected by various kinds of bias,and the popularity bias is one of the important factors affecting the accuracy of recommendation system.The traditional debias method is mainly based on item feature,and in-troduces penalty factor to suppress the recommendation of popular items,without considering the influence of user interest and time.To solve this problem,we propose an Adaptive Popularity and Dynamic Interest(APDI)method based on item popularity and user dynamic interest.The main factors affecting the popularity bias are comprehensively analyzed from the two aspects of item popularity and user per-sonalization,and according to the different sensitivity of item quality,conformity effect and user interest to time,the comprehensive score of item popularity and user dynamic interest within the corresponding time interval is calculated to reduce the popularity bias more effectively.By using the exponential weighted moving average method,the user interest score is calculated according to the influence of time decay degree on the user's current interest,so as to capture the user's short-term interest preference.The validity of the proposed method is verified on three real datasets.Experimental results show that the proposed method can effectively improve the precision,recall and normalized discounted cumulative gain of recommendation.
万方数据
维普
