Clustering-based diversity improvement in top-N recommendation

The major aim of recommender algorithms has been to predict accurately the rating value of items. However, it has been recognized that accurate prediction of rating values is not the only requirement for achieving user satisfaction. One other requirement, which has gained importance recently, is the diversity of recommendation lists. Being able to recommend a diverse set of items is important for user satisfaction since it gives the user a richer set of items to choose from and increases the chance of discovering new items. In this study, we propose a novel method which can be used to give each user an option to adjust the diversity levels of their own recommendation lists. Experiments show that the method effectively increases the diversity levels of recommendation lists with little decrease in accuracy. Compared to the existing methods, the proposed method, while achieving similar diversification performance, has a very low computational time complexity, which makes it highly scalable and allows it to be used in the online phase of the recommendation process.     

Knowledge distillation meets recommendation: collaborative distillation for top-N recommendation

Knowledge and Information Systems - Knowledge distillation (KD) is a successful method for transferring knowledge from one model (i.e., teacher model) to another model (i.e., student model)....     

一种基于随机游走的歌曲推荐算法

推荐系统可以帮助人们在海量的数据中发现所需的有价值的信息。传统的协同过滤推荐算法根据历史数据中用户对项目的各种行为操作构建用户-项目评分矩阵,进而计算相似度,从而预测用户对项目的偏好程度进行推荐。但因为评分数据通常较为稀疏,使得推荐的准确性不高,从而不能很好地对用户进行推荐。针对这个问题,提出一种结合场论理论的随机游走歌曲推荐算法,融合歌曲评分相似度和歌曲基本信息相似度,降低歌曲间综合相似度矩阵的稀疏性,并将物理学中的场论理论和歌曲的重要度结合,构造转移概率矩阵,从而实现歌曲推荐。实验表明,该算法较协同过滤算法的推荐准确性更佳。     

A movie recommendation algorithm based on genre correlations

Since the late 20th century, the number of Internet users has increased dramatically, as has the number of Web searches performed on a daily basis and the amount of information available. A huge amount of new information is transferred to the Web on a daily basis. However, not all data are reliable and valuable, which implies that it may become more and more difficult to obtain satisfactory results from Web searches. We often iterate searches several times to find what we are looking for. To solve this problem, researchers have suggested the use of recommendation systems. Instead of searching for the same information several times, a recommendation system proposes relevant information. In the Web 2.0 era, recommendation systems often rely on collaborative filtering by users. In general, a collaborative filtering approach based on user information such as gender, location, or preference is effective. However, the traditional approach can fail due to the cold-start problem or the sparsity problem, because initial user information is required for this approach to be effective. Recently, several attempts have been made to tackle these collaborative filtering problems. One such attempt used category correlations of contents. For instance, a movie has genre information provided by movie experts and directors. This category information is more reliable than user ratings. Moreover, newly created content always has category information, allowing avoidance of the cold-start problem. In this study, we consider a movie recommendation system and improve the previous algorithms based on genre correlations to correct its shortcomings. We also test the modified algorithm and analyze the results with respect to two characteristics of genre correlations.     

Leveraging pointwise prediction with learning to rank for top-N recommendation

World Wide Web - Pointwise prediction and Learning to Rank (L2R) are two hot strategies to model user preference in recommender systems. Currently, these two types of approaches are often...     

基于双维度云模型的协同过滤推荐算法

随着网络的迅速普及,网络应用多样化趋势加快,大数据时代已经到来,电商业对于在线推荐系统的要求也越来越高。为了满足人们的需求,传统的推荐算法必须得到改进和发展。本文主要考虑数据比较稀疏及数据规模较大的情况下如何更好地实现协同过滤推荐算法,针对这两个方面的不足,结合Hadoop平台的优势,基于双维度云模型的协同过滤推荐算法由此应运而生了。通过实验,利用云模型和双维度数据,有关数据稀疏性问题得到了合理的解决,预测准确率方面也得到了一定的改进。并且通过MapReduce处理流程,使大数据环境下运行困难效率低下的缺陷得到了弥补。     

一种基于用户兴趣分析的协同过滤推荐算法

传统的协同过滤推荐算法以用户对所有物品的评分向量作为计算用户相似度的依据,没有考虑到物品属性对用户兴趣的反映。为此,提出一种新的改进的相似度计算方法,引入了"用户兴趣分布矩阵"的定义,设计了启发式的评分预测方式,即根据兴趣相似度选出TOP-K用户之后,以用户标记的物品数量作为该用户的权重来预测评分。在Movielens数据集上的测试结果表明,改进后的算法相比传统的算法在平均绝对误差(MAE)上降低了7.3%。     

基于深度学习的混合推荐算法

针对传统的推荐系统存在推荐精度较低且冷启动较严重的问题,综合考虑评论文本与评分而提出改进的稀疏边缘降噪自动编码与近邻项目影响力的矩阵分解模型相结合的混合推荐方法。通过改进的稀疏边缘降噪自动编码模型从商品评论文本中来提取商品特征,将用户评论与评分联合,同时综合考虑了近邻用户对于目标用户的影响,将近邻项目影响力整合到矩阵分解算法之中,与传统的协同深度学习模型(CDL)和混合SDAE模型相比,最高可提升8. 370%。     

基于概念分层的个性化推荐算法

协同过滤算法(couaborative filtering)目前较为成功地应用于个性化推荐系统中,但随着系统规模的扩大,面临很严重的稀疏性问题,制约了推荐效果。文中提出概念分层的方法对用户-项矩阵进行改进,同时使用交易数据和点击流数据,将相似用户选择项与多层次关联规则推荐项相结合,在稀疏数据集上表现出较好的性能。     

基于CoP建模的协作过滤推荐方法

传统的协作过滤推荐方法主要基于个人兴趣特征来实现推荐。在组织内部协作场景下,为实现知识共享与重用,推荐系统不仅要考虑用户兴趣,还应考虑用户和用户组的任务。传统的协作过滤推荐方法已不能满足要求。CoP是组织内部人员管理的主要形式,它的特征是其成员任务的反映。基于已有的协作过滤推荐研究与D-S理论,提出了一种CoP特征构建算法,并以此为基础研究了面向CoP的协作过滤推荐。     

基于人工鱼群算法的协同过滤推荐算法

基于原始人工鱼群算法,提出在觅食行为中保留较优值以替代随机值,在追尾和聚群行为中比较最优值和中心值再作移动行为的选择,在迭代进行中,实现视野的自适应调整.这样改进后的人工鱼群算法应用于协同过滤推荐系统中,实现用户聚类,从而提高协同过滤推荐系统的最近邻查询速度,降低搜索开销.实验测试结果显示了改进的人工鱼群算法具有收敛速度快,稳定性高的特性,且能获得较优的聚类目标值.将改进的人工鱼群算法用于协同过滤推荐算法中,提高了算法的推荐精度.     

A new slope one based recommendation algorithm using virtual predictive items

Although the Slope One family of algorithms provides an appealing solution to the scalability problem in collaborative filtering recommendation systems, the data sparsity problem as a major issue still remains open. Many of the recent algorithms rely on sophisticated methods which not only have negative effect on the scalability of Slope One, but also need some additional information extra to ratings matrix. To address these problems in this paper, we have proposed a novel method based on Weighted Slope One algorithm which introduces virtual predictive items in relatively sparse ratings databases. These virtual items are those which neither have rated by active users nor have deviation to active items. The strength of our approach lies in its ability to manage the data sparsity problem without using any extra information. Indeed, it uses the ratings data which are common in collaborative filtering systems. Our proposed algorithm is scalable, easy to implement and updatable on the fly (without changing comprehensively). Experimental results on the MovieLens and Netflix datasets show the effectiveness of the proposed algorithm in handling data sparsity problem. It also outperforms some state-of-the-art collaborative filtering algorithms in terms of prediction quality.     

基于协同过滤推荐的电价计划智能推荐算法

在智慧电网中,电力公司可以主动推荐定制的售电方案给潜在用户,但现有的推荐算法存在着精确度不高、方案不合理等缺点.为解决以上问题,基于协同过滤策略,开发一种电力计划推荐方案.通过提供一些容易获得的家电产品数据,对居民用户进行不同方案的预测评级,为用户选择合适的方案和合理的电价.在实验阶段,通过不同的数值实验评价该方法的性能,实验结果表明,EPR算法在推荐结果的准确性上优于其它策略.     

基于重要性排名的协同过滤推荐算法

传统的协同过滤忽略系统中不同用户和条目的重要性对推荐结果的影响.针对此问题,提出了一种基于用户和条目重要性的改进协同过滤算法,该算法将条目的重要性融合到用户相似性的度量方法中,将用户的重要性融入到预测评分的计算方法中;为度量系统中每个条目和用户的重要性,提出了ItemRank和UserRank算法.在MovieLens数据集上的实验结果表明,提出的算法可以显著提高推荐系统的推荐质量.     

Interactive resource recommendation algorithm based on tag information

With the popularization of social media and the exponential growth of information generated by online users, the recommender system has been popular in helping users to find the desired resources from vast amounts of data. However, the cold-start problem is one of the major challenges for personalized recommendation. In this work, we utilized the tag information associated with different resources, and proposed a tag-based interactive framework to make the resource recommendation for different users. During the interaction, the most effective tag information will be selected for users to choose, and the approach considers the users’ feedback to dynamically adjusts the recommended candidates during the recommendation process. Furthermore, to effectively explore the user preference and resource characteristics, we analyzed the tag information of different resources to represent the user and resource features, considering the users’ personal operations and time factor, based on which we can identify the similar users and resource items. Probabilistic matrix factorization is employed in our work to overcome the rating sparsity, which is enhanced by embedding the similar user and resource information. The experiments on real-world datasets demonstrate that the proposed algorithm can get more accurate predictions and higher recommendation efficiency.     

基于广义Choquet积分的职位推荐算法

为解决传统职位推荐算法推荐结果范围小,无法对跨行就业率高的人群进行有效推荐的问题,提出基于人口统计学数据及广义Choquet积分的职位推荐算法。利用层次分析法构建用户信息层次结构模型,结合广义Shapley函数与离散Choquet积分,通过计算最优相似人群进行职位推荐。与传统模型实验对比,推荐模型在雷达招聘网站提供的数据集上准确率提高了8.35%,有效验证了其可行性和有效性。     

基于知识图谱的用户偏好推荐算法

针对现有的基于知识图谱推荐算法中,缺乏与用户之间的交互,忽略物品间连接关系的问题,提出了基于知识图谱的用户偏好推荐算法.首先,为了更准确地获得用户对物品的偏好类型,增加知识图谱与用户信息的交互.其次,为知识图谱引入注意力机制,使节点传播中关注相似度更高的节点,并在递归传播中优化了节点的嵌入,将两部分关系权重图叠加得到对...     

基于贝叶斯理论的协同过滤推荐算法

考虑到在协同过滤算法中邻居集合的有效性是影响推荐质量的重要因素,提出了基于贝叶斯理论的协同过滤推荐方法,该方法利用贝叶斯理论分析用户对项目特征值的喜好度.在计算相似度时,考虑用户喜好度,在此基础上计算目标项目的最近邻居.实验结果表明该算法可以提高推荐系统的推荐质量.