摘要
协同过滤算法是推荐系统中使用最广泛的算法之一,随着个性化推荐技术的发展,传统的协同过滤算法在数据稀疏的情况下推荐的准确率较低,同时没有考虑用户的兴趣会随着时间的推移发生动态变化等因素,传统的协同过滤推荐算法已无法满足个性化推荐的需求。论文针对以上问题提出一种融合算法,将K-means算法和隐语义模型相结合,提出基于用户聚类和时间隐语义模型的推荐算法K-T-LFM(K-means algorithm clustering users and Time Based Latent Factor Model)。该算法根据用户的属性特征,采用最大-最小准则确定初始质心的K-means算法将用户聚类,解决了新用户登录的冷启动问题,降低了矩阵的稀疏程度和矩阵规模;根据艾宾浩斯遗忘曲线提出时间函数,并融合传统隐语义模型对聚类中的用户评分稀疏矩阵进行填充,有效缓解了数据的稀疏性,同时考虑了时间因素对用户的兴趣偏好的影响,提高了推荐算法的准确性。通过MovieLens数据集进行实验对比,该算法较其他的协同过滤算法准确率有所提升。
Collaborative filtering algorithm is one of the most widely used algorithms in the recommendation system.With the development of personalized recommendation technology,the accuracy of traditional collaborative filtering algorithm is low in the case of sparse data,and it does not consider the dynamic changes of users'interests over time.The traditional collaborative filtering recommendation algorithm can not meet the needs of personalized recommendation.To solve the above problems,this paper propos⁃es a fusion algorithm,which combines K-means algorithm with Latent Factor Model,and proposes a recommendation algorithm K-T-LFM(K-means algorithm clustering users and Time Based Latent Factor Model).The algorithm uses the maximum-minimum criterion to determine the initial centroid K-means algorithm to cluster users according to the attributes of users,solves the cold start problem of new user logins,and reduces the sparseness and scale of the matrix.A time function is proposed based on the Ebb⁃inghaus forgetting curve,and the traditional implicit semantic model is integrated to fill the sparse matrix of user ratings in the clus⁃ter,the influence of time factor on user's interest preference is considered.Through the MovieLens data set for experimental compari⁃son,the accuracy of this algorithm is improved compared with other collaborative filtering algorithms.
作者
吴祺
聂文惠
WU Qi;NIE Wenhui(School of Computer and Communication Engineering,Jiangsu University,Zhenjiang 212000)
出处
《计算机与数字工程》
2023年第3期561-565,共5页
Computer & Digital Engineering
关键词
协同过滤
用户聚类
时间隐语义模型
推荐
collaborative filtering
user clustering
time based latent factor model
recommendation
