Item Recommendation Research Articles

The impact of off-hour mechanical thrombectomy therapy on outcomes for acute ischemic stroke: A systematic review and meta-analysis

BackgroundThe impact of off-hours and on-hours mechanical thrombectomy (MT) treatment for acute ischemic stroke (AIS) is not well understood. We conducted a systematic review and met-analysis comparing outcomes between patients undergoing off-hours MT versus on-hours MT. MethodsThis study is reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses recommendations. A systematic literature review of the English language literature was conducted using Web of Science, Embase, Scopus, and PubMed databases. We included all original studies comparing off-hour and on-hour outcomes or time metrics in AIS undergoing MT. R software version 4.3.1 and the ‘meta’ statistical package were used to analyze all the data presented in this study. ResultsWe included 26 studies with 82,850 patients. Patients undergoing MT during off-hours achieved lower rates of 90-day functional independence (OR 0.92, 95% CI 0.86–0.99; p = 0.04) and successful recanalization (OR 0.89, 95% CI 0.81–0.98; p = 0.014). Patients undergoing off-hours MT experienced similar rates of 90-day mortality (OR 1.07, 95% CI 0.96–1.19; p = 0.21) and sICH (OR 1.04, 95% CI 0.85–1.28; p = 0.674). Patients in the off-hour group experienced longer onset to door time (MD = 12.23 min; 95% CI 4.53–19.93; p = 0.002), imaging to puncture time (MD = 10.59 min; 95% CI 4.00–17.19; p = 0.002), and door to recanalization time (MD = 13.31 min; 95% CI 4.60–22.03; p = 0.003). ConclusionsPatients undergoing MT for AIS during off-hours experienced lower rates of functional independence. This may be attributed to treatment delays during off-hours. Future studies should work to optimize hospital workflows and identify factors which may contribute to treatment delays.

Improved conversational recommender system based on dialog context

Abstract Conversational recommender system (CRS) needs to be seamlessly integrated between the two modules of recommendation and dialog, aiming to recommend high-quality items to users through multiple rounds of interactive dialogs. Items can typically refer to goods, movies, news, etc. Through this form of interactive dialog, users can express their preferences in real time, and the system can fully understand the user’s thoughts and recommend corresponding items. Although mainstream dialog recommendation systems have improved the performance to some extent, there are still some key issues, such as insufficient consideration of the entity’s order in the dialog, the different contributions of items in the dialog history, and the low diversity of generated responses. To address these shortcomings, we propose an improved dialog context model based on time-series features. Firstly, we augment the semantic representation of words and items using two external knowledge graphs and align the semantic space using mutual information maximization techniques. Secondly, we add a retrieval model to the dialog recommendation system to provide auxiliary information for generating replies. We then utilize a deep timing network to serialize the dialog content and more accurately learn the feature relationship between users and items for recommendation. In this paper, the dialog recommendation system is divided into two components, and different evaluation indicators are used to evaluate the performance of the dialog component and the recommendation component. Experimental results on widely used benchmarks show that the proposed method is effective.

Intent with knowledge-aware multiview contrastive learning for recommendation

User–item interactions on e-commerce platforms involve various intents, such as browsing and purchasing, which require fine-grained intent recognition. Existing recommendation methods incorporate latent intent into user–item interactions; however, they overlook important considerations. First, they fail to integrate intents with semantic information in knowledge graphs, neglecting intent interpretability. Second, they do not exploit the structural information from multiple views of latent intents in user–item interactions. This study established the intent with knowledge-aware multiview contrastive learning (IKMCL) model for explanation in recommendation systems. The proposed IKMCL model converts latent intent into fine-grained intent, calculates intent weights, mines latent semantic information, and learns the representation of user–item interactions through multiview intent contrastive learning. In particular, we combined fine-grained intents with a knowledge graph to calculate intent weights and capture intent semantics. The IKMCL model performs multiview intent contrastive learning at both coarse-grained and fine-grained levels to extract semantic relationships in user–item interactions and provide intent recommendations in structural and semantic views. In addition, an intent-relational path was designed based on multiview contrastive learning, enabling the capture of semantic information from latent intents and personalized item recommendations with interpretability. Experimental results using large benchmark datasets indicated that the proposed model outperformed other advanced methods, significantly improving recommendation performance.

Unveiling the Wave: Unraveling Sales Trends and Unlocking Market Basket Insights

This article presents sales trends and market basket analysis for Food and Beverages (F&B) items recommendation using linear regression time series and association rules. The purpose of this study is to help businesses to make better informed decisions by understanding their customers’ purchasing behaviors and the sales trends of the F&B items. This includes the study of both cross- selling and up-selling patterns using market basket analysis. To improve the market basket analysis, food name clustering has been carried out to group those foods having similarity ratio higher or equal to 80 based on the Levenshtein distances between the food names. This study analyzes sales data from a mobile app platform that register the Malaysian F&B hawkers (or merchants) and app users (or buyers) who sell and purchase F&B respectively. This study is based on sales trends analysis and association rules of product recommendations by examining FP-Growth and the relationship between product items bought. This study provides valuable insights for local Malaysian hawkers who are looking to improve their F&B items recommendations and to make better data-driven decisions while planning their marketing activities. It helps to improve sales by either up or cross-sell their products more effectively.

A Reinforcement Learning List Recommendation Model Fused with Graph Neural Networks

Existing list recommendation methods present a list consisting of multiple items for feedback recommendation to user requests, which has the advantages of high flexibility and direct user feedback. However, the structured representation of state data limits the embedding of users and items, making them isolated from each other, missing some useful infomation for recommendation. In addition, the traditional non-end-to-end learning series takes a long time and accumulates errors. During the model training process, the results of each task can easily affect the next calculation, thus affecting the entire training effect. Aiming at the above problems, this paper proposes a Reinforcement Learning List Recommendation Model Fused with a Graph Neural Network, GNLR. The goal of this model is to maximize the recommendation effect while ensuring that the list recommendation system accurately analyzes user preferences to improve user experience. To this end, firstly, we use an user–item bipartite graph and Graph Neural Network to aggregate neighborhood information for users and items to generate graph structured representation; secondly, we adopt an attention mechanism to assign corresponding weights to neighborhood information to reduce the influence of noise nodes in heterogeneous information networks; finally, we alleviate the problems of traditional non-end-to-end methods through end-to-end training methods. The experimental results show that the method proposed in this paper can alleviate the above problems, and the recommendation hit rate and accuracy rate increase by about 10%.

A Multimodal User-Adaptive Recommender System

Traditional recommendation systems have predominantly relied on user-provided ratings as explicit input. Concurrently, visually aware recommender systems harness inherent visual cues within data to decode item characteristics and deduce user preferences. However, the untapped potential of incorporating item images into the recommendation process warrants investigation. This paper introduces an original convolutional neural network (CNN) architecture that leverages multimodal information, connecting user ratings with product images to enhance item recommendations. A central innovation of the proposed model is the User-Adaptive Filtering Module, a dynamic component that utilizes user profiles to generate personalized filters. Through meticulous visual influence analysis, the effectiveness of these filters is demonstrated. Furthermore, experimental results underscore the competitive performance of the approach compared to traditional collaborative filtering methods, thereby offering a promising avenue for personalized recommendations. This approach capitalizes on user adaptation patterns, enhancing the understanding of user preferences and visual attributes.

TIEN: Temporal interest-aware evolution model for “Next Item Recommendation”

The “next item recommendation” is used to predict the next item that a user may be interested in and its main challenges is identifying the user's interests. The current state-of-the-art methods mostly adopt time-series-based recommendation methods but ignore the role of time intervals in the interest evolution process. This motivated us to propose a new Time Interest-aware Evolutionary Network model (TIEN) to predict the next item that a user may be interested in. TIEN highlights the role of time intervals in the interest evolution process and divides interests into long-term interests and short-term interests. In TIEN: (1) We devise a time interval-based attention network TAN to capture the user's interests, which employs item embeddings, behavior sequence embeddings, and time intervals to calculate the user's interest attention; (2) We devise an ATT-GRU network based on candidate activation vectors to capture the interest evolution process, which skillfully leverages the reset gate, hidden layer state, item information, time intervals, etc. to participate in interest evolution; (3) We introduce a dual-network collaborative training mechanism, where the obtained interests of users are treated as weights to guide the interest evolution process. Experimental results on multiple benchmark datasets show that TIEN has higher performance than the current state-of-the-art methods.

A Neural Inference of User Social Interest for Item Recommendation

AbstractUser-generated content is daily produced in social media, as such user interest summarization is critical to distill salient information from massive information for recommendation tasks. While the interested messages (e.g., tags or posts) from a single user are usually sparse becoming a bottleneck for existing methods, we propose a neural inference method (NIGraphNet) by mining user social interest for item recommendation. It can unearth user latent topics combined with user relation learning. Specifically, we exploit a neural variational inference approach to learn the distributions between user interests and hidden topics. (We denote it as interest-topic distributions in the following.) Then, we adopt a unified graph-based training loss that jointly learns the hidden topics and user relations for item recommendation. Experiments on two datasets collected from well-known social media platforms demonstrate the superior performance of our model in the tasks of user interest summarization and item recommendation. Further discussions also show that exploiting the latent topic representations and user relations is conducive to the user’s automatic language understanding.

Privacy-preserving graph convolution network for federated item recommendation

In traditional recommender systems, we often build models based on a centralized storage of user data, which however will lead to user privacy concerns and risks. In this paper, we study an emerging and important recommendation problem called federated item recommendation (FIR), in which a recommendation model is built with decentralized data of user-item interactions in a privacy-aware manner, i.e., the personal behavior data of each user does not leave the owner. Recently, graph neural network (GNN) has been widely recognized as a state-of-the-art solution for item recommendation with implicit feedback since it is able to model the high-order connectivity between users and items. However, it is very challenging to exploit the high-order connectivity information in a decentralized user-item interaction graph without compromising user privacy. To address that, we propose a GNN-based federated recommendation framework, i.e., privacy-preserving graph convolution network (P-GCN), for the studied problem of FIR. Our P-GCN can leverage the high-order connectivity information like a centralized GCN model such as LightGCN, though it is built using a decentralized user-item graph. To achieve that, we design a novel privacy-preserving graph convolution approach based on secure aggregation and employ item-based user representation to compensate for the performance loss it causes due to the protection of user privacy. Moreover, we improve a group-wise concealing strategy for protecting user privacy. Empirical studies on three datasets show that our P-GCN can achieve similar or even better performance comparing with the non-federated (i.e., centralized) counterpart, and outperforms all the existing federated methods for the studied problem.

Multi-aspect features of items for time-ordered sequential recommendation

The key to sequential recommendation modeling is to capture dynamic users’ interests. Existing sequential recommendation methods (e.g., self-attention mechanism) have achieved extraordinary success in modeling users’ interests. However, these models ignore that users have different levels of preferences for different aspects of items, failing to capture users’ most concerning aspects. In addition, they are highly dependent on the quality of training data, which may lead to overfitting of the model when the training data is insufficient. To address the above issues, we propose a novel sequence-aware model (Multi-Aspect Features of Items for Time-Ordered Sequential Recommendation, MFITSRec), which combines the features of items with user behavior sequences to learn more complex item-item and item-attribute relationships. Moreover, the model uses a self-attention network based on an absolute time relationship, which can better represent the changes in users’ interests and capture users’ preferences for particular aspects of items. Extensive experiments on five datasets demonstrate that our model outperforms various baseline models. In particular, the model’s prediction accuracy has been significantly improved on sparse datasets.

Transfer Learning across Graph Convolutional Networks: Methods, Theory, and Applications

Graph neural networks have been widely used for learning representations of nodes for many downstream tasks on graph data. Existing models were designed for the nodes on a single graph, which would not be able to utilize information across multiple graphs. The real world does have multiple graphs where the nodes are often partially aligned . For examples, knowledge graphs share a number of named entities though they may have different relation schema; collaboration networks on publications and awarded projects share some researcher nodes who are authors and investigators, respectively; people use multiple web services, shopping, tweeting, rating movies, and some may register the same email account across the platforms. In this paper, we propose partially aligned graph convolutional networks to learn node representations across the models. We provide multiple methods such as model sharing, regularization, and alignment reconstruction, as well as theoretical analysis to positively transfer knowledge across the set of partially aligned nodes. Extensive experiments on real-world knowledge graphs, collaboration networks, and bipartite rating graphs show the superior performance of our proposed methods on relation classification, link prediction, and item recommendation.

FASTER: A Dynamic Fairness-assurance Strategy for Session-based Recommender Systems

When only users’ preferences and interests are considered by a recommendation algorithm, it will lead to the severe long-tail problem over items. Therefore, the unfair exposure phenomenon of recommended items caused by this problem has attracted widespread attention in recent years. For the first time, we reveal the fact that there is a more serious unfair exposure problem in session-based recommender systems (SRSs), which learn the short-term and dynamic preferences of users from anonymous sessions. Considering the fact that in SRSs, recommendations are provided multiple times and item exposures are accumulated over interactions in a session, we define new metrics both for the fairness of item exposure and recommendation quality among sessions. Moreover, we design a dynamic F airness- A ssurance ST rategy for s E ssion-based R ecommender systems ( FASTER ). FASTER is a post-processing strategy that tries to keep a balance between item exposure fairness and recommendation quality. It can also maintain the fairness of recommendation quality among sessions. The effectiveness of FASTER is verified on three real-world datasets and five original algorithms. The experiment results show that FASTER can generally reduce the unfair exposure of different session-based recommendation algorithms while still ensuring a high level of recommendation quality.

A Comprehensive Empirical Study of Heterogeneity in Federated Learning

Federated learning (FL) is becoming a popular paradigm for collaborative learning over distributed, private datasets owned by non-trusting entities. FL has seen successful deployment in production environments, and it has been adopted in services such as virtual keyboards, auto-completion, item recommendation, and several IoT applications. However, FL comes with the challenge of performing training over largely heterogeneous datasets, devices, and networks that are out of the control of the centralized FL server. Motivated by this inherent challenge, we aim to empirically characterize the impact of device and behavioral heterogeneity on the trained model. We conduct an extensive empirical study spanning nearly 1.5K unique configurations on five popular FL benchmarks. Our analysis shows that these sources of heterogeneity have a major impact on both model quality and fairness, causing up to 4.6× and 2.2× degradation in the quality and fairness, respectively, thus shedding light on the importance of considering heterogeneity in FL system design.

Best Practice Recommendations for Dysphagia Management in Intensive Care Unit (ICU): A Delphi Study from Multidisciplinary Experts in Turkey.

There is no study about all aspects of oropharyngoesophageal (OPE) dysphagia from diagnosis to follow-up in a multidisciplinary manner in the world. In order to close this gap, we aimed to create a recommendation study that can be used in clinical practice, addressing all aspects of dysphagia in the ICU in detail with the opinion of experienced multidisciplinary experts. This recommendation paper was generated by a multidisciplinary team, using the seven-step process and a three-modified Delphi round via e-mail. Firstly, 15 open-ended questions were created, and then detailed recommendations including general principles, management, diagnosis, rehabilitation, and follow-up were created with the answers from these questions, Each recommendation item was voted on by the experts as overall consensus (strong recommendation), approaching consensus (weak recommendation), and divergent consensus (not recommended).In the first Delphi round, a questionnaire consisting of 413 items evaluated with a scale of 0-10 was prepared from the opinions and suggestions given to 15 open-ended questions. In the second Delphi round, 55.4% were accepted and revised suggestions were created. At the end of the third Delphi round, the revised suggestion form was approved again and the final proposals containing 133 items were created. This study includes comprehensive and detailed recommendations, including a broad perspective from diagnosis to treatment and follow-up, as detailed as possible, for management of dysphagia in patients with both oropharyngeal- and esophageal-dysphagia in ICU.

Influence Maximization in Complex Networks by Using Evolutionary Deep Reinforcement Learning

Influence maximization (IM) in complex networks tries to activate a small subset of seed nodes that could maximize the propagation of influence. The studies on IM have attracted much attention due to their wide applications such as item recommendation, viral marketing, information propagation and disease immunization. Existing works mainly model the IM problem as a discrete optimization problem, and use either approximate or meta-heuristic algorithms to address this problem. However, these works are hard to find a good tradeoff between effectiveness and efficiency due to the NP-hard and large-scale network properties of the IM problem. In this article, we propose an evolutionary deep reinforcement learning algorithm (called EDRL-IM) for IM in complex networks. First, EDRL-IM models the IM problem as a continuous weight parameter optimization of deep Q network (DQN). Then, it combines an evolutionary algorithm (EA) and a deep reinforcement learning algorithm (DRL) to evolve the DQN. The EA simultaneously evolves a population of individuals, and each of which represents a possible DQN and returns a solution to the IM problem through a dynamic markov node selection strategy, while the DRL integrates all information and network-specific knowledge of DQNs to accelerate their evolution. Systematic experiments on both benchmark and real-world networks show the superiority of EDRL-IM over the state-of-the-art IM methods in finding seed nodes.

User Interest Based Contextual Conversational Recommender System

The Conversational recommender system is for item recommendation during human-machine dialogue, where items can be movies, music, and goods. Scenarios can be intelligent customer service and shopping assistants. The system's goal is to generate high-quality recommendations in short-round interactions. There are two main problems with the system. Firstly, recommending through dialogues is a hard problem as less contextual information is available, resulting in fewer signals for recommendation and less precision for user interest modeling. Secondly, the recommender system considers the absence of interaction between a user and an item as a negative sample. However, this approach can introduce bias in the objectives because non-interaction could potentially be a positive sample. To tackle the previously mentioned concerns, this paper suggests a new model. This model adds an extra knowledge graph to enrich contextual information. Then, this model introduces contrastive learning to separate the distribution between positive and negative samples and improve learning efficiency. This paper further tests the model on the ReDial dataset, and experiments have shown that the method is effective and performs better than previous work.

KGTN: Knowledge Graph Transformer Network for explainable multi-category item recommendation

Most existing methods for recommendation assume that all the items are of the same category. However, single-category recommendations are no longer sufficient to fulfill the diverse needs of users in the real world. Existing methods cannot be directly applied to recommend multiple categories of items due to the different properties and complex relationships among them. However, a drawback of these approaches is their assumption of a fixed and homogeneous data format, making them incapable of addressing the challenge of recommending multiple categories of items. Recommendation of multiple categories of items to users concurrently has become a challenging task. To tackle this problem, we design a novel method Knowledge Graph Transformer Network (KGTN) for explainable multi-category item recommendation, inspired by advances in knowledge graph in the field of recommendation. Knowledge graph and neural network methods have shown advantages in addressing recommendation problems in heterogeneous graph structures. This is because different categories of items have unique attributes and dimensions, which can be effectively represented and integrated by incorporating knowledge graph and graph neural network techniques. That is, our approach can handle heterogeneous collaborative knowledge graphs composed of users and items, and can mine hidden path relationships without defining original paths. In addition, its algorithmic process has interpretability. Specifically, graph transformer layer converts the heterogeneous input graph into a useful meta-path graph for recommended task; graph convolution layer learns the node representation on the new meta-path graph. Finally, we calculate the inner product of user and item vector representations to output the probability for the recommendation. At the same time, we use the critical path in the learned useful meta-path graph as the explanation. Comprehensive experiments on three datasets demonstrate that KGTN achieves state-of-the-art performance over existing baselines.

A patent recommendation method based on KG representation learning

The number of patents is constantly increasing, providing an invaluable knowledge base. But retrieving this knowledge becomes ever more challenging with the increase in data. In this context, this study proposes a patent recommendation method based on knowledge graph (KG) representation. Although KG-based recommendation systems received broad attention in the literature, there is still a lack of literature that explores how it can be used for patent recommendations. First, and in terms of KG construction, the existing methods encounter issues when it comes to simultaneously accounting for accuracy and recall rate. This study proposes a combination of rule-based and deep learning methods for KG construction to solve this problem. Second, existing methods have performance and semantic association limitations when learning KG representations. This study therefore uses high-performance graph neural networks for encoding and models the semantic association limitations in graphs using traditional KG representation learning methods. Finally, existing recommendation systems usually focus on the semantic correlation between interaction items in recommendations, ignoring their direct relationships. This study therefore applies the KG representation learning method for the patent recommendation task, using KG representation learning to model the semantic association and constraints in the product–patent heterogeneous graph. Experiments based on a practical case example demonstrate that the proposed method outperforms baseline methods, and that it can guide innovative design processes. Findings have important implications for research and practice.

Sequential ensemble learning for next item recommendation

Predicting the next item that users may engage in is a key task of recommender systems, and many methods have been proposed to deal with this task from different aspects. Theoretically, the proper ensemble of multiple different methods (a.k.a. base models) can make more accurate and stable recommendations. However, most of the existing ensemble methods rely on static aggregation strategies, which fail to capture base models’ dynamic predictive abilities for each user over time. In addition, most of the existing diversity measures used in regression or classification ensemble methods rely on a distance metric of base models’ outputs, which makes it intractably apply for next-item recommendation whose base models output sequential ranking lists.To solve the above problems, we propose a Sequential Ensemble Method, named SEM, to aggregate different base models for next-item recommendation. We assume users’ concentration and base models’ expertise can be inferred from users’ sequential behaviors and base models’ prediction results. Therefore, we propose to explicitly model base models’ dynamic predictive abilities on different users over time based on users’ concentration and base models’ expertise. In addition, we propose a new diversity measure for sequential ranking ensemble, which can perform diversity-based learning over time for better performance of next-item recommendation. Extensive experiments on six real-world data sets show that our method consistently outperforms state-of-the-art methods.

Cold-start item recommendation for representation learning based on heterogeneous information networks with fusion side information

In recommendation systems, the cold-start problem refers to the situation that users or items have too few interaction events with the system to be recommended, such as clicks or ratings. The cold-start problem leads to the loss of new users and difficulty in recommending new items and weakens the performance of recommendations. In this paper, we focus on alleviating the cold-start problem of items. Previous work has used the structure of item homogeneous graphs to establish connections between items and generate item embeddings. However, the homogeneous graph of a single node cannot fully explore the potential structural features between users and goods. Due to the flexibility of heterogeneous data, recommendation systems can also employ the Heterogeneous Information Network(HIN) to characterize complex, heterogeneous data. In this paper, a novel Heterogeneous Graph Embedding Learning based Skip-Gram (HSG) model is proposed to mitigate the cold-start problem. First, we introduce three different types of subnetworks and combine them to construct heterogeneous graphs. Then meaningful sequences of nodes are generated via meta-path-based random walk for network embedding. Finally, a Skip-Gram model incorporating side information is used to learn item representations. The learned node embeddings are used as pre-training for the Matrix Factorization (MF) model. The HSG model is proven to be effective by extensive experiments on four publicly available datasets.