Socialized healthcare service recommendation using deep learning

Socialized recommender system recommends reliable healthcare services for users. Ratings are predicted on the healthcare services by merging recommendations given by users who has social relations with the active users. However, existing works did not consider the influence of distrust between users. They recommend items only based on the trust relations between users. We therefore propose a novel deep learning-based socialized healthcare service recommender model, which recommends healthcare services with recommendations given by recommenders with both trust relations and distrust relations with the active users. The influences of recommenders, considering both the node information and the structure information, are merged via the deep learning model. Experimental results show that the proposed model outperforms the existing works on prediction accuracy and prediction coverage simultaneously, even for cold start users or users with very sparse trust relations. It is also computational less expensive.

     

设备端基于深度学习的智能家居服务推荐框架

随着智能家居的普及,用户期望通过自然语言指令实现智能设备的控制,并希望获得个性化的智能家居服务。然而,现有的挑战包括智能设备的互操作性和对用户环境的全面理解。针对上述问题,提出一个支持设备端用户智能家居服务推荐个性化的框架。首先,构建智能家居的运行时知识图谱,用于反映特定智能家居中的上下文信息,并生成用例场景语句;其次,利用预先收集的通用场景下,用户的自然语言指令和对应的用例场景语句训练出通用推荐模型;最后,用户在设备端以自然语言管理智能家居设备和服务,并通过反馈微调通用模型的权重得到个人模型。在基本指令集、复述集、场景指令集三个数据集上的实验表明,用户的个人模型相比于词嵌入方法的准确率提升了6.5%~30%,与Sentence-BERT模型相比准确率提升了2.4%~25%,验证了设备端基于深度学习的智能家居服务框架具有较高的服务推荐准确率,能够有效地管理智能家居设备和服务。     

Location-based deep factorization machine model for service recommendation

Applied Intelligence - The era of everything as a service led to an explosion of services with similar functionalities on the internet. Quickly obtaining a high-quality service has become a...     

基于云‒端融合的个性化推荐服务系统

主流个性化推荐服务系统通常利用部署在云端的模型进行推荐,因此需要将用户交互行为等隐私数据上传到云端,这会造成隐私泄露的隐患。为了保护用户隐私,可以在客户端处理用户敏感数据,然而,客户端存在通信瓶颈和计算资源瓶颈。针对上述挑战,设计了一个基于云?端融合的个性化推荐服务系统。该系统将传统的云端推荐模型拆分成用户表征模型和排序模型,在云端预训练用户表征模型后,将其部署到客户端,排序模型则部署到云端;同时,采用小规模的循环神经网络（RNN）抽取用户交互日志中的时序信息来训练用户表征,并通过Lasso算法对用户表征进行压缩,从而在降低云端和客户端之间的通信量以及客户端的计算开销的同时防止推荐准确率的下跌。基于RecSys Challenge 2015数据集进行了实验,结果表明,所设计系统的推荐准确率和GRU4REC模型相当,而压缩后的用户表征体积仅为压缩前的34.8%,计算开销较低。     

Web服务个性化推荐研究综述

随着Web服务的广泛使用和互联网上服务数量的增加,如何向用户提供最佳的服务选择列表成为了新的挑战.Web服务个性化推荐实现了由被动接受用户请求向主动感知用户需求的转变.个性化的Web服务推荐方法已经成为Web服务发现和选择的有效辅助手段.Web服务的个性化推荐技术也成为了近年来服务计算领域的研究热点.对当前Web服务个性化推荐的文献进行了归类分析,总结了当前Web服务个性化推荐的技术现状、研究方法和实验的数据集,列出了未来Web服务个性化推荐研究热点和挑战.     

Deep knowledge-aware framework for web service recommendation

The Journal of Supercomputing - Web services are products in the era of service-oriented computing and cloud computing. Considering the information overload problem arising from the task of...     

Context-aware QoS prediction for web service recommendation and selection

QoS prediction is one of the key problems in Web service recommendation and selection. The context information is a dominant factor affecting QoS, but is ignored by most of existing works. In this paper, we employ the context information, from both the user side and service side, to achieve superior QoS prediction accuracy. We propose two novel prediction models, which are capable of using the context information of users and services respectively. In the user side, we use the geographical information as the user context, and identify similar neighbors for each user based on the similarity of their context. We study the mapping relationship between the similarity value and the geographical distance. In the service side, we use the affiliation information as the service context, including the company affiliation and country affiliation. In the two models, the prediction value is learned by the QoS records of a user (or a service) and the neighbors. Also, we propose an ensemble model to combine the results of the two models. We conduct comprehensive experiments in two real-world datasets, and the experimental results demonstrate the effectiveness of our models.     

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.     

移动电话内容服务系统的个性化推荐

移动电话内容服务系统允许移动用户通过移动互联技术浏览、购买和下载系统内容,是当前移动增值领域研究的热点。具有较强的时空灵活性,但在信息浏览、查找方面存在明显的局限性。提出了一个基于移动电话内容服务系统的个性化推荐系统．介绍了从寻找目标用户到实现推荐的全过程。实验结果表明。所介绍的个性化推荐系统可以有助于解决内容服务系统用户访问受限、资源迷茫的问题。     

A graph-based QoS prediction approach for web service recommendation

Applied Intelligence - With the development of the Internet, the recommendation based on Quality of Service(QoS) is proven to be an efficient way to deal with the ever-increasing web services in...     

An integrated model based on deep multimodal and rank learning for point-of-interest recommendation

Modeling point-of-Interest (POI) for recommendations is vital in location-based social networks, yet it is a challenging task due to data sparsity and cold-start problems. Most existing approaches incorporate content features into a probabilistic matrix factorization model using unsupervised learning, which results in inaccuracy and weak robustness of recommendations when data is sparse, and the cold-start problems remain unsolved. In this paper, we propose a deep multimodal rank learning (DMRL) model that improves both the accuracy and robustness of POI recommendations. DMRL exploits temporal dynamics by allowing each user to have time-dependent preferences and captures geographical influences by introducing spatial regularization to the model. DMRL jointly learns ranking for personal preferences and supervised deep learning models to create a semantic representation of POIs from multimodal content. To make model optimization converge more rapidly while preserving high effectiveness, we develop a ranking-based dynamic sampling strategy to sample adverse or negative POIs for model training. We conduct experiments to compare our DMRL model with existing models that use different approaches using two large-scale datasets obtained from Foursquare and Yelp. The experimental results demonstrate the superiority of DMRL over the other models in creating cold-start POI recommendations and achieving excellent and highly robust results for different degrees of data sparsity.

     

知识表示学习研究进展

新型深度学习推荐模型已广泛应用至现代推荐系统,其独有的特征——包含万亿嵌入参数的嵌入层,带来的大量不规则稀疏访问已成为模型预估的性能瓶颈. 然而,现有的推荐模型预估系统依赖CPU对内存、外存等存储资源上的嵌入参数进行访问,存在着CPU-GPU通信开销大和额外的内存拷贝2个问题,这增加了嵌入层的访存延迟,进而损害模型预估的性能. 提出了一种基于GPU直访存储架构的推荐模型预估系统GDRec.GDRec的核心思想是在嵌入参数的访问路径上移除CPU参与,由GPU通过零拷贝的方式高效直访内外存资源. 对于内存直访,GDRec利用统一计算设备架构（compute unified device architecture,CUDA）提供的统一虚拟地址特性,实现GPU 核心函数（kernel）对主机内存的细粒度访问,并引入访问合并与访问对齐2个机制充分优化访存性能;对于外存直访,GDRec实现了一个轻量的固态硬盘（solid state disk,SSD）驱动程序,允许GPU从SSD中直接读取数据至显存,避免内存上的额外拷贝,GDRec还利用GPU的并行性缩短提交I/O请求的时间. 在3个点击率预估数据集上的实验表明,GDRec在性能上优于高度优化后的基于CPU访存架构的系统NVIDIA HugeCTR,可以提升多达1.9倍的吞吐量.

     

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%。     

基于主题模型的个性化图书推荐算法

针对传统推荐算法精准度不高的问题,在潜在狄利克雷分布(LDA)主题挖掘模型的基础上提出了一种新的适用于图书推荐(BR)的数据挖掘模型——BR_LDA模型。通过对目标借阅者的历史借阅数据与其他图书数据进行内容相似度分析,得到与目标借阅者历史借阅图书内容相似度较高的其他图书。通过对目标借阅者的历史借阅数据及其他借阅者的历史借阅数据进行相似性分析,得到最近邻借阅者的历史借阅数据。通过求解图书被推荐的概率,最终得到目标借阅者潜在感兴趣的图书。特别地,当推荐数量为4000时,BR_LDA模型比基于多特征方法和关联规则方法精准度分别提高了6.2%、4.5%;当推荐数量为500时,BR_LDA模型比协同过滤的近邻方法和矩阵分解方法分别提高了2.1%、0.5%。实验表明本模型能够更准确地向目标借阅者推荐历史感兴趣类别的新图书及潜在感兴趣的新类别的图书。     

An intelligent web service group-based recommendation system for long-term composition

A modern model long-term composed service (LCS) with a group recommendation system has an indefinite lifespan. An LCS is used as a long-term business goal, and for a business committed to its customers, support will be provided to customers enabling them to book, e.g. an automotive service through online web services by providing information that the LCS then uses to offer more support. However, identifying the exact service to meet the user requirement is essential. Service composition has been identified as the key task in achieving various QoS performances. There exist various approaches that involve service composition according to the throughput and popularity. However, they fail to achieve the expected performance. Towards improving the performance of the LCS, a novel LCS that is based on the user queries of a group of persons is developed to give the best business services based on previous travel details and services. The method carries out service selection and composition according to the ratings provided by users towards any service. Additionally, the method considers the user-to-service rating and service-to-service rating, which are measured according to the coupling quality. Therefore, the proposed novel LCS provides better services based on the user ratings for particular business queries. The method ranks the services according to the rating values to perform service composition, with consideration of the detection of similar user groups and utilization of the rating values in service selection. We aim to propose a novel LCS work based on group ratings and a group of services. This work is intended to reduce the time complexity of changes in the LCS network using the group recommendation system.

     

基于图神经网络与深度学习的商品推荐算法

基于图神经网络的推荐算法可以提取传统方法无法提取用户与商品之间的关联关系.目前此类算法大多忽略了用户和商品的评论数据中所存在的一般偏好.针对这一问题,提出了一种方法,在利用图神经网络提取关联关系的同时,利用深度学习提取评论的优势提取用户和商品的一般偏好,并进行特征融合来提升推荐效果.在四组公共数据集中进行了对比实验,使用召回率和归一化折损累计增益作为评价指标,并通过消融实验验证了方法的有效性.实验表明该方法比已有相关算法的效果更好.两种网络的特征融合对推荐效果有提升作用.     

Adaptive recommendation for photo pose via deep learning

Multimedia Tools and Applications - With the development of image acquisition devices and the popularity of smart phones, more and more people would like to upload their photos to diverse social...     

QoS感知的Web服务推荐模型*

服务注册中心往往推荐给用户一组等价服务,且服务的QoS来自于不同用户的反馈,如何根据用户的偏好和上下文环境推荐最合适的服务在SOA中非常重要。提出了一个QoS感知的Web服务推荐模型,该模型通过感知用户上下文克服了传统方法带来的用户QoS体验与服务注册中心QoS信息不一致的问题,使用层次分析法获得用户QoS偏好,综合评估备选服务的QoS性能,给出最适合当前用户的服务。通过实验说明了用户上下文感知的合理性,并给出了一个服务推荐的例子。     

基于深度学习的个性化对话内容生成方法

人机对话系统是人机交互领域一个非常重要的研究方向,开放域聊天机器人的研 究受到了广泛关注。现有的聊天机器人主要存在 3 个方面的问题：①无法有效捕捉上下文情境 信息,导致前后对话内容缺乏逻辑关联。②大部分不具备个性化特征,导致聊天过程千篇一律, 且前后对话内容可发生矛盾。③倾向于生成“我不知道”、“对不起”等无意义的通用回复内容, 极大降低用户的聊天兴趣。本研究中利用基于 Transformer 模型的编解码(Encoder-Decoder)结构 分别构建了通用对话模型和个性化对话模型,通过编码历史对话内容和个性化特征信息,模型 可以有效捕捉上下文情境信息以及个性化信息,实现多轮对话过程,且对话内容符合个性化特 征。实验结果表明,基于 Transformer 的对话模型在困惑度(perplexity)和 F1 分数评价指标上相 比于基线模型得到了一定的提升,人工评价显示模型可以正常进行多轮交互对话过程,生成内 容多样性高,且符合给定的个性化特征。