基于人群移动模式先验的兴趣点推荐

兴趣点推荐是基于位置的社交网络中的一项重要任务，为用户提供个性化的地点推荐。然而，当前的兴趣点推荐方法主要学习用户在兴趣点上的签到历史和用户间的社交关系网络，城市人群出行规律无法得到有效利用。首先提出了人群移动模式提取框架(Human Mobility Pattern Extraction, HMPE),利用图神经网络作为人群移动模式的提取器，引入注意力机制捕获城市交通模式的时空信息。HMPE通过制定下游任务，设计上采样模块将表征向量还原为任务目标，实现端到端的框架学习训练，完成人群移动模式提取器的预训练。其次，提出了兴趣点推荐算法HMRec(Human Mobility Recommendation),引入了人群移动模式的先验知识，使得推荐结果更符合城市中的人类出行意愿。对比实验结果显示，HMRec的表现优于基线模型。最后，讨论了兴趣点推荐存在的问题和未来的研究方向。     

基于多线程并行强化学习的数据库索引推荐

建立索引是提高数据库性能的一个重要方法。目前随着强化学习算法的发展,出现了一系列使用强化学习解决索引推荐问题（index selection problem,ISP）的方法。针对现有的深度强化学习索引推荐算法训练时间长,训练不够稳定的问题,提出了一个基于A2C的索引推荐算法PRELIA。该算法加入负载索引扫描行数特征矩阵,并对奖励值进行归一化处理,旨在提高索引选择的准确性和效率,减少索引空间占用。在不同数据集上的实验结果表示,该算法可以在保证与比较的算法相当的索引推荐质量同时,推荐出的索引占用更小的存储空间,同时训练时间比基线算法时间提高了4倍以上。     

对话推荐算法研究综述

推荐系统是一种通过理解用户的兴趣和偏好帮助用户过滤大量无效信息并获取感兴趣的信息或者物品的信息过滤系统.目前主流的推荐系统主要基于离线的、历史的用户数据,不断训练和优化线下模型,继而为在线的用户推荐物品,这类训练方式主要存在3个问题:基于稀疏且具有噪声的历史数据估计用户偏好的不可靠估计、对影响用户行为的在线上下文环境因素的忽略和默认用户清楚自身偏好的不可靠假设.由于对话系统关注于用户的实时反馈数据,获取用户当前交互的意图,因此“对话推荐”通过结合对话形式与推荐任务成为解决传统推荐问题的有效手段.对话推荐将对话系统实时交互的数据获取方式应用到推荐系统中,采用了与传统推荐系统不同的推荐思路,通过利用在线交互信息,引导和捕捉用户当前的偏好兴趣,并及时进行反馈和更新.在过去的几年里,越来越多的研究者开始关注对话推荐系统,这一方面归功于自然语言处理领域中语音助手以及聊天机器人技术的广泛使用,另一方面受益于强化学习、知识图谱等技术在推荐策略中的成熟应用.将对话推荐系统的整体框架进行梳理,将对话推荐算法研究所使用的数据集进行分类,同时对评价对话推荐效果的相关指标进行讨论,重点关注于对话推荐系统中的后...     

基于多任务自编码器的MOOC课程推荐模型

为解决在线学习当中,学习者行为的数量远少于在线课程的样本总数所产生的数据稀疏问题,提出一种基于多任务自编码器的课程推荐模型（multi-task autoencoder course recommendation model,MAEM）。通过分析学习者的学习行为,将总体任务分为两个子任务：任务一是学习者浏览课程章节列表行为,任务二是完成课程50%的学习行为,通过共享网络底部的隐藏层提高泛化能力。模型总体划分为共享嵌入、自编码器与分解预测、任务组合3个模块,3个模块协同工作,旨在突破训练数据稀疏问题。将MAEM与7种常用的推荐算法比较,实验结果表明,MAEM算法优于7种热门的推荐算法,验证了其在课程推荐中的有效性。     

中文微博的Hashtag话题相关性分析

Hashtag(微博话题词)是发布者为微博信息创建的话题标签,能帮助用户在海量微博数据中高效发现热点话题。Hashtag由用户创建的特性使得不同的Hashtag可能代表着同一个话题,挖掘Hashtag之间的话题相关性将有助于热点话题发现和聚合展示。研究了Hashtag之间相关性分析问题,抽取了Hashtag文本特征、微博内容、Hashtag的出现次数-时间分布以及Hashtag共现等一系列特征,以分析Hashtag之间的话题相关性。在新浪微博数据上的实验结果显示,这一系列特征组合能较好地帮助Hashtag相关性分析。     

课程资源的融合知识图谱多任务特征推荐算法

在采集在线学习信息时,普遍存在数据缺失的情况,使得课程资源推荐时可能因数据稀疏导致推荐的效果不理想。为了解决上述问题,基于端对端的深度学习框架,提出了融合知识图谱的多任务特征推荐算法（Multi-Layer Knowledge graph Recommendation,MLKR）。基于多任务特征学习,在任务中嵌入知识图谱;在任务之间通过交叉压缩单元建立潜在特征和实体之间的高阶联系,从而建立推荐模型。实现了基于学习者目标、兴趣、知识水平的课程资源精准推荐。实验结果表明,MLKR推荐算法训练时长和预测准确率均优于基于用户或物品的协同过滤算法和逻辑回归模型,在课程资源推荐领域具有一定的应用价值。     

融合知识图谱和短期偏好的推荐算法

近年来,将知识图谱作为辅助信息来增强推荐越来越受到研究者的关注.由于知识图谱学习任务的目标是还原知识图谱中三元组的关系,并非是以推荐任务为目标,导致了知识图谱学习任务很难高效地帮助推荐任务提升推荐性能.另外,用户兴趣易被短期的环境和心情所影响.针对以上两点,提出了一种融合了知识图谱信息和短期偏好的推荐模型(MKASR).首先,通过RippleNet算法提取用户和知识图谱实体的关系对,然后将这些关系对按照知识图谱三元组的形式存储和参与训练;采用基于注意力机制的双向GRU网络从用户近期交互的物品序列中提取用户的短期偏好;其次,采用多任务学习的方法同时训练知识图谱学习模块和推荐模块,并得到用户和物品的特征表示;最后,通过这些特征表示和用户的短期偏好向用户综合推荐.在真实数据集MovieLens-1M和Book-Crossing上进行实验,采用AUC、ACC、Precision和Recall指标进行评估,实验结果表明,提出的模型优于其他的基准模型.     

小波卷积增强的对比学习推荐算法

推荐算法是一种用于解决信息过载问题的方法,序列化推荐通过建模用户购买的物品序列预测下一个物品。现有的序列化推荐算法通常忽视用户行为序列中的噪声、跨序列信息和物品间的组合依赖等问题,导致推荐性能受限。为此,提出一种小波卷积增强的对比学习推荐算法WCLR。利用数据的内在相关性获得自监督信号,并根据预训练的方法来增强数据表示。给出3个辅助的自监督学习任务,利用信息最大化原理学习属性、物品、序列与邻居序列的相关性,通过互信息最大化提供一种统一的方式描述不同类型数据间的相关性。由于小波卷积网络能提取物品的组合依赖,降低用户交互序列中的噪声,设计一个多核小波卷积模块,通过多尺寸用户序列多方面捕获用户的潜在兴趣,将自监督学习和小波卷积融入到推荐算法模型中,降低序列数据稀疏性和噪声,提高推荐精度。在LastFM、Beauty和Toys 3个数据集上的实验结果表明,与8个序列化推荐模型相比,WCLR算法的命中率、归一化折损累计增益和平均倒数秩分别提升了3.30%、1.47%和2.17%。     

融合异构图注意力网络的知识推荐

为解决在线课程平台直接推荐服务忽视用户知识水平差异的问题,本文提出一种知识推荐方法。通过应用异构信息网络和注意力机制,将知识推荐问题视为强化学习任务,模拟用户与知识概念之间的互动,并利用异构信息网络学习用户的语义表示,通过元路径引导用户偏好的传播。实验结果表明,相较于其他方法,该方法在性能上表现得更为优越。     

基于知识图谱和图嵌入的个性化学习资源推荐

面对海量的在线学习资源,学习者往往面临“信息过载”和“信息迷航”等问题,帮助学习者高效准确地获取适合自己的学习资源来提升学习效果,已成为研究热点.针对现有方法存在的可解释性差、推荐效率和准确度不足等问题,提出了一种基于知识图谱和图嵌入的个性化学习资源推荐方法,它基于在线学习通用本体模型构建在线学习环境知识图谱,利用图嵌入算法对知识图谱进行训练,以优化学习资源推荐中的图计算效率.基于学习者的学习风格特征进行聚类来优化学习者的资源兴趣度,以获得排序后的学习资源推荐结果.实验结果表明,相对于现有方法,所提方法能在大规模图数据场景下显著提升计算效率和个性化学习资源推荐的准确度.     

Hashtag Recommendation Based on Multi-Features of Microblogs

Hashtag recommendation for microblogs is a very hot research topic that is useful to many applications involving microblogs. However, since short text in microblogs and low utilization rate of hashtags will lead to the data sparsity problem, it is difficult for typical hashtag recommendation methods to achieve accurate recommendation. In light of this, we propose HRMF, a hashtag recommendation method based on multi-features of microblogs in this article. First, our HRMF expands short text into long text, and then it simultaneously models multi-features (i.e., user, hashtag, text) of microblogs by designing a new topic model. To further alleviate the data sparsity problem, HRMF exploits hashtags of both similar users and similar microblogs as the candidate hashtags. In particular, to find similar users, HRMF combines the designed topic model with typical user-based collaborative filtering method. Finally, we realize hashtag recommendation by calculating the recommended score of each hashtag based on the generated topical representations of multi-features. Experimental results on a real-world dataset crawled from Sina Weibo demonstrate the effectiveness of our HRMF for hashtag recommendation.     

Hashtag Recommendation Approach Based on Content and User Characteristics

ABSTRACT

Twitter has become a popular microblogging service that allows millions of active users share news, emergent social events, personal opinions, etc. That leads to a large amount of data producing every day and the problem of managing tweets becomes extremely difficult. To categorize the tweets and make easily in searching, the users can use the hashtags embedding in their tweets. However, valid hashtags are not restricted which lead to a very heterogeneous set of hashtags created on Twitter, increasing the difficulty of tweet categorization. In this paper, we propose a hashtag recommendation method based on analyzing the content of tweets, user characteristics, and currently popular hashtags on Twitter. The proposed method uses personal profiles of the users to discover the relevant hashtags. First, a combination of tweet contents and user characteristics is used to find the top-k similar tweets. We exploit the content of historical tweets, used hashtags, and the social interaction to build the user profiles. The user characteristics can help to find the close users and enhance the accuracy of finding the similar tweets to extract the hashtag candidates. Then a set of hashtag candidates is ranked based on their popularity in long and short periods. The experiments on tweet data showed that the proposed method significantly improves the performance of hashtag recommendation systems.     

Combining long-term and short-term user interest for personalized hashtag recommendation

Hashtags, terms prefixed by a hash-symbol #, are widely used and inserted anywhere within short messages (tweets) on micro-blogging systems as they present rich sentiment information on topics that people are interested in. In this paper, we focus on the problem of hashtag recommendation considering their personalized and temporal aspects. As far as we know, this is the first work addressing this issue specially to recommend personalized hashtags combining longterm and short-term user interest.We introduce three features to capture personal and temporal user interest: 1) hashtag textual information; 2) user behavior; and 3) time. We offer two recommendation models for comparison: a linearcombined model, and an enhanced session-based temporal graph (STG) model, Topic-STG, considering the features to learn user preferences and subsequently recommend personalized hashtags. Experiments on two real tweet datasets illustrate the effectiveness of the proposed models and algorithms.     

Phrase-based hashtag recommendation for microblog posts

In microblogs, authors use hashtags to mark keywords or topics. These manually labeled tags can be used to benefit various live social media applications (e.g., microblog retrieval, classification). However, because only a small portion of microblogs contain hashtags, recommending hashtags for use in microblogs are a worthwhile exercise. In addition, human inference often relies on the intrinsic grouping of words into phrases. However, existing work uses only unigrams to model corpora. In this work, we propose a novel phrase-based topical translation model to address this problem. We use the bag-of-phrases model to better capture the underlying topics of posted microblogs. We regard the phrases and hashtags in a microblog as two different languages that are talking about the same thing. Thus, the hashtag recommendation task can be viewed as a translation process from phrases to hashtags. To handle the topical information of microblogs, the proposed model regards translation probability as being topic specific. We test the methods on data collected from realworld microblogging services. The results demonstrate that the proposed method outperforms state-of-the-art methods that use the unigram model.     

标签推荐方法研究综述

随着互联网信息的爆炸式增长,标签(由用户指定用来描述项目的关键词)在互联网信息检索领域中变得越来越重要.为在线内容赋予合适的标签,有利于更高效的内容组织和内容消费.而标签推荐通过辅助用户进行打标签的操作,极大地提升了标签的质量,标签推荐也因此受到了研究者们的广泛关注.总结出标签推荐任务的三大特性,即项目内容的多样性、标...     

基于图对比学习的MOOC推荐方法

随着MOOC在线教育平台的飞速发展,课程和用户数量激增,学习者在面对种类繁多的课程时往往较难选择,传统的推荐方法在MOOC课程推荐中应用存在对曝光次数较低的课程推荐效果差和对噪声数据鲁棒性不足的问题。为给学习者提供高质量的推荐,提出一种图对比学习的MOOC推荐方法,同时针对二分图结构给出一种新的数据增强方法。对输入的用户项目交互的二分图随机添加或者删除边进行数据增强,得到两个子视图,使用图卷积神经网络对原始二分图和两个子视图进行节点特征提取得到用户和项目的节点表征,并构建推荐监督任务和对比学习的辅助任务进行联合优化,在此基础上将用户和项目的节点表征进行点积获得推荐结果。在MOOC数据集上进行Top-K推荐的实验结果表明,相较于LightGCN模型,该方法在Recall@5和NDCG@5上均有显著提升,最高分别提升7.8%和7.3%,能够有效提高模型对于曝光次数较低的课程的推荐准确性和对于噪声数据的鲁棒性。     

Sequence recommendation based on deep learning

In order to solve the cold start problem of traditional recommendation algorithm, the sequence change of user interaction information and deep learning are gradually considered as a key feature of commodity recommendation system. However, most of the existing recommendation methods based on the sequence changes assume that all the interaction information of users is equally important for recommendation, which is not always applicable in real scenarios, because the interaction process of user items is full of randomness and contingency. In this article, we study how to reduce the randomness and contingency between session sequences, make full use of the association between session sequences in the interaction process of users by Deep Learning. In order to better simulate the change of session sequence in the real scene, we adopt sequence sampling methods to transform the single classification problem into sequence modeling problem. And attention mechanism is added to reduce the interference of the recommendation model in the sequence due to the contingency and randomness of the user in the shopping. Finally, through the verification of real data, the MRR@20 index of the improved model is 20% higher than the benchmark level.