摘要
在个性化推荐算法中,相似度计算方法是决定推荐算法准确度的关键因素之一。在MovieLens数据集下,传统相似度的计算没有考虑用户共同评分项目、用户评分时间及用户平均评分的影响,以至于在数据稀疏的情况下不能很好地度量用户间的相似度,导致推荐结果不准确,因此,就以上问题,提出加入三个修正因子改进传统的相似度计算方法。在MovieLens数据集下,实验结果表明,改进后的协同过滤算法的平均绝对误差(MAE)比基于改进相似性度量的项目协同过滤推荐算法(ICF_IPSS)低7.4%,比基于加权多融合偏好与结构相似度的协同过滤算法(MCF)低6%,比协同过滤的相似度融合改进算法(ICF_SI)低1%,可见改进的算法在推荐准确性方面有显著的提高。
In personalized recommendation algorithm, similarity computation is one of the key factors to decide the accuracy of recommendation algo- rithm. In the MovieLens data set, the traditional calculation does not consider the common user similarity score scores of time and user items, the average score, that cannot be a good measure of similarity between users in the case of sparse data, leading to the recommenda- tion result is not accurate, because of this, the above problems, adds three improved correction factor the traditional similarity calculation method. In the MovieLens data set, the experimental results show that the average absolute error of the improved collaborative filtering algo- rithm （MAE） than the improved similarity measure project collaborative filtering recommendation algorithm based on （ICF_IPSS） 7.4% low- er than the weighted fusion preference and structural similarity in collaborative filtering algorithm （MCF） is 6% lower than the similarity of cooperation filtering fusion algorithm （ICF_SI） low 1%, visible the improved algorithm has a significant improvement in the accuracy of rec- ommendation.
作者
高兴前
王晓峰
GAO Xing-qian;WANG Xiao-feng(College of Information Engineering,Shanghai Maritime University,Shanghai 201306)
出处
《现代计算机》
2018年第17期31-36,共6页
Modern Computer
关键词
协同过滤
推荐算法
Person相似度
共同评分项目
Collaborative Filtering
Recommendation Algorithm
Person Similarity
Common Scoring Project
