Recommender systems help users find items they are likely to interact within the near future, such as products to buy in e-commerce or songs to play in music websites. The Traditional recommendation ...methods make predictions based on long-term user profiles, i.e., the items a user interacted with in the past while ignoring the time and order of the interactions. Recent findings, however, suggest that users may exhibit interest to items in a certain order depending on situations and more recent items in a sequence have a larger impact on the subsequent choices. Moreover, in many practical applications, user-item interactions are organized into short sessions, where each session reflects the user's short-term interest in addition to long-term preferences. Leveraging both long-term user profiles and short-term sequential patterns from sessions can lead to more accurate models known as the session-aware recommendation methods. In this paper, we explore various strategies to integrate user long-term preferences with session patterns encoded by recurrent neural networks (RNNs). The strategies include integrating user embeddings with input and output of session RNNs, integrating with fixed or adaptive contributions of the user and session components by using a specially designed gating mechanism. We conducted an empirical evaluation of three publicly available datasets. The results indicate that combining user long-term profiles with the output of session RNNs yields improved predictions and the proposed adaptive integration model outperforms the state-of-the-art sequential and session-aware recommendation methods.
With wide use of cloud computing technologies, microblog is used more widely for services providing more personal communities by user information sharing, dissemination and acquisition. In Microblog ...environment, hashtag is used to find messages with a specific theme or content, which can greatly facilitate information diffusion, microblog searching, event detection and topic analysis, etc. Recommending relevant hashtags to users in the cloud is challenging, because hashtags are created at tremendous speed alongside microblogs, and scattered in micro-blogging systems without a systematic organization. In this paper, a personalized hashtag recommendation approach is proposed according to the latent topical information in microblogs. With users represented by user-topics distribution, the proposed approach finds top-k similar users, then computes all hashtags’ frequencies appeared in these users, and finally the most relevant hashtags are recommended to user. In order to excavate latent topical information, a Latent Dirichlet Allocation (LDA)-based topic model is also proposed, named Hashtag-LDA, which can greatly enhance the influence of hashtags on latent topics’ generation by jointly modeling hashtags and words in microblogs. Hashtag-LDA can not only find meaningful latent topics, but also find global hashtags and the relationships between topics and hashtags. The experimental results on real Twitter dataset show that the proposed recommendation approach outperforms the related methods and Hashtag-LDA is effective.
•Hashtag-LDA recommendation approach finds more relevant users in microblogs.•Combine user profile-based collaborative and LDA-based collaborative filtering.•Model the relations between users, hashtags and words through latent topics.
Collaborative filtering (CF) is one of the most popular recommendation methods, and the co-rating-based similarity measurement is widely used in CF for predicting ratings of unfamiliar items. In ...addition to rating information, social trust has now been considered useful in collaborative recommendations. In this work, we present a hybrid approach that combines user ratings and social trust for making better recommendations. In contrast to other trust-aware recommendation works, our approach exploits distrust links and investigates their propagation effects. In addition, our approach combines the k-nearest neighbors and the matrix factorization methods to maximize the advantages of both rating and trust information. Several series of experiments are conducted, in which different types of social trust are incrementally included to evaluate the presented approach. The results show that distrust information is beneficial in ratings prediction, and the developed hybrid approach can effectively enhance the recommendation performance.
Many e-commerce sites present additional item recommendations to their visitors while they navigate the site, and ample evidence exists that such recommendations are valuable for both customers and ...providers. Academic research often focuses on the capability of recommender systems to help users discover items they presumably do not know yet and which match their long-term preference profiles. In reality, however, recommendations can be helpful for customers also for other reasons, for example, when they remind them of items they were recently interested in or when they point site visitors to items that are currently discounted. In this work, we first adopt a systematic statistical approach to analyze what makes recommendations effective in practice and then propose ways of operationalizing these insights into novel recommendation algorithms. Our data analysis is based on log data of a large e-commerce site. It shows that various factors should be considered in parallel when selecting items for recommendation, including their match with the customer’s shopping interests in the previous sessions, the general popularity of the items in the last few days, as well as information about discounts. Based on these analyses we propose a novel algorithm that combines a neighborhood-based scheme with a deep neural network to predict the relevance of items for a given shopping session.
Currently, artificial intelligence (AI) recommendations are widely used to alleviate the phenomenon of information overload, and how to enhance the effectiveness of AI recommendations is a very ...important issue. Based on theory of uses and gratifications, this paper applied neuroscience technology of event-related potential (ERP) to investigate how different AI recommendation methods (explicit and implicit) and product types (similar and related) affect consumers’ decision-making process and neuropsychology mechanisms. Behavioral results showed that consumers were more likely to accept the implicit recommendations when recommending similar products. However, when recommending related products, consumers were more willing to accept explicit recommendations. At the neural level, ERP results provided underlying cognitive evidence for exploring consumers’ decision-making on AI recommendations. There was a two-stage cognitive process of consumers on different AI recommendation methods and product types. In the early cognitive stage, a greater P2 amplitude was elicited by recommendation of similar products than that of related products, reflecting an automatic and primary attention allocation process. In the later cognitive stage, the recommendation method of implicit than that of explicit evoked a larger P3 amplitude when recommending similar products, while the recommendation method of explicit than that of implicit induced a greater P3 amplitude when recommending related products, reflecting an advanced categorization evaluation process. These findings have important theoretical and practical implications for gaining a deeper understanding of consumers’ decision making on AI recommendations and promoting the development of AI recommendations.
Sequential recommendation problems have received increased research interest in recent years. In such scenarios, the task is to suggest items to users to consume next, given their past interaction ...history, e.g., the next movie to watch or the next item to place in the shopping cart. A number of machine learning models were proposed recently for the task of sequential recommendation, with the latest ones based on deep learning techniques, in particular on Transformers. Given the often surprisingly competitive performance of simpler nearest-neighbor methods for the related problem of session-based recommendation, we investigate the use of nearest-neighbor methods for sequential recommendation problems. Our analysis on four datasets shows that nearest-neighbor methods achieve comparable or better performance than the recent Transformer-based bert4rec method on two of them. However, the deep learning method outperforms the simple methods for the two larger datasets, confirming previous hypotheses that neural methods work best when more data is available. As a further result of our experiments, we found additional evidence that sampled metrics must be used with care, as they may not be predictive of an algorithm ranking that would be observed with the non-sampled, full evaluation.
Abstract Background Guideline developers can: 1) adopt existing recommendations from others; 2) adapt existing recommendations to their own context; or 3) create recommendations de novo . Monetary ...and non-monetary resources, credibility, maximization of uptake, as well as logical arguments should guide the choice of the approach and processes. Objective To describe a potentially efficient model for guideline production based on adoption, adaptation and/or de novo development of recommendations utilizing the GRADE Evidence to Decision (EtD) frameworks. Study Design and Setting We applied the model in a new national guideline program producing 22 practice guidelines. We searched for relevant evidence that informs the direction and strength of a recommendation. We then produced GRADE EtDs for guideline panels to develop recommendations. Results We produced a total of 80 EtD frameworks in approximately 4 months and 146 EtDs in approximately 6 months in two waves. Use of the EtD frameworks allowed panel members understand judgments of others about the criteria that bear on guideline recommendations, and then make their own judgments about those criteria in a systematic approach. Conclusion The “GRADE-ADOLOPMENT” approach to guideline production combines adoption, adaptation, and, as needed, de novo development of recommendations. If developers of guidelines follow EtD criteria more widely and make their work publically available, this approach should prove even more useful.
Since December 2019, an epidemic caused by novel coronavirus (2019-nCoV) infection has occurred unexpectedly in China. As of 8 pm, 31 January 2020, more than 20 pediatric cases have been reported in ...China. Of these cases, ten patients were identified in Zhejiang Province, with an age of onset ranging from 112 days to 17 years. Following the latest
National recommendations for diagnosis and treatment of pneumonia caused by 2019-nCoV
(the 4th edition) and current status of clinical practice in Zhejiang Province, recommendations for the diagnosis and treatment of respiratory infection caused by 2019-nCoV for children were drafted by the National Clinical Research Center for Child Health, the National Children’s Regional Medical Center, Children’s Hospital, Zhejiang University School of Medicine to further standardize the protocol for diagnosis and treatment of respiratory infection in children caused by 2019-nCoV.
Crowdsourcing is a distributed computing paradigm that utilizes human intelligence or resources from a crowd of workers. Existing solutions of task recommendation in crowdsourcing may leak private ...and sensitive information about both tasks and workers. To protect privacy, information about tasks and workers should be encrypted before being outsourced to the crowdsourcing platform, which makes the task recommendation a challenging problem. In this paper, we propose a privacy-preserving task recommendation scheme (PPTR) for crowdsourcing, which achieves the task-worker matching while preserving both task privacy and worker privacy. In PPTR, we first exploit the polynomial function to express multiple keywords of task requirements and worker interests. Then, we design a key derivation method based on matrix decomposition, to realize the multi-keyword matching between multiple requesters and multiple workers. Through PPTR, user accountability and user revocation are achieved effectively and efficiently. Extensive privacy analysis and performance evaluation show that PPTR is secure and efficient.
Mining potential and valuable medical knowledge from massive medical data to support clinical decision-making has become an important research field. Personalized medicine recommendation is an ...important research direction in this field, aiming to recommend the most suitable medicines for each patient according to the health status of the patient. Personalized medicine recommendation can assist clinicians to make clinical decisions and avoid the occurrence of medical abnormalities, so it has been widely concerned by many researchers. Based on this, this paper makes a comprehensive review of personalized medicine recommendation. Specifically, we first make clear the definition of personalized medicine recommendation problem; then, starting from the key theories and technologies, the personalized medicine recommendation algorithms proposed in recent years are systematically classified (medicine recommendation based on multi-disease, medicine recommendation with combination pattern, medicine recommendation with additional knowledge, and medicine recommendation based on feedback) and in-depth analyzed; and this paper also introduces how to evaluate personalized medicine recommendation algorithms and some common evaluation indicators; finally, the challenges of personalized medicine recommendation problem are put forward, and the future research direction and development trends are prospected.