| 摘 要: 基于距离模型的协同过滤通过计算用户间已知评分的距离，并使用该距离来预测目标用户的未知评分，但该类算法因预测需要使用所有邻居而导致需要大量缓存距离计算结果。针对这一问题，设计了一种融合用户相似性与用户评分距离的个性化推荐算法，基于用户间的相似性对邻居进行筛选，使用筛选之后的邻居集合预测未知评分。基于MovieLens数据与现有几种经典算法进行比较实验，证明了设计方法的有效性，在降低29%邻居数量的基础上，该算法提高了预测准确性、推荐列表排序性能等多个关键指标。 |
| 关键词: 推荐系统；协同过滤；相似性；评分距离；个性化推荐 |
|
中图分类号: TP182
文献标识码: A
|
| 基金项目: 国家自然科学基金项目(61803264). |
|
| A Personalized Recommendation Algorithm Integrating User Similarity and Rating Distance |
|
SU Zhan, HUANG Zhong, AI Jun
|
(School of Optical -Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China)
suzhan@foxmail.com; huang970622@126.com; aijun@outlook.com
|
| Abstract: Collaborative filtering based on distance model calculates the distance of known ratings between users and uses this distance to predict the unknown ratings of target users, but this kind of algorithm requires a large amount of cached distance calculation results due to the need to use all neighbors for prediction. Aiming at this problem, this paper proposes to design a personalized recommendation algorithm integrating user similarity and user rating distance, which filters neighbors based on the similarity between users, and uses the filtered neighbor set to predict unknown ratings. Comparison experiment between MovieLens data and several existing classical algorithms verifies the effectiveness of the proposed algorithm. Upon reducing the number of neighbors by 29%, the proposed algorithm improves the prediction accuracy, recommendation list sorting performance and other key indicators. |
| Keywords: recommendation system; collaborative filtering; similarity; rating distance; personalized recommendation |
