User Clustering Research Articles

MOREM: An evolutionary multitasking optimization algorithm for multi-objective recommendations

Multi-objective evolutionary algorithms (MOEAs) have been successfully applied in recommender systems, where the recommendation is modeled as a multi-objective optimization problem by considering both accuracy and non-accuracy metrics. However, to recommend to all users, the population individuals in most of the exiting MOEA-based recommendation algorithms are usually encoded in the form of matrix encoding, and with the increased number of users and items, it will lead to a large search space. To this end, in this paper, we propose a novel evolutionary multitasking algorithm MOREM to tackle the challenge of multi-objective recommendations, where a related auxiliary task is constructed by both user and item reduction to help solving the complex original task by knowledge transfer. Specifically, a task generation strategy for creating auxiliary task is firstly proposed to cluster users by user-rating information (i.e., user reduction), where only a part of important items (i.e., item reduction) are recommended for the central user within each cluster with the aim to greatly reduce the size of the matrix encoding. In addition, a novel knowledge transfer mechanism is proposed, which can effectively achieve knowledge migration between the two tasks. Experimental results on real-world datasets show that MOREM outperforms several state-of-the-art algorithms for multi-objective recommendations.

Optimizing resource allocation for cluster D2D-assisted fog computing networks: A three-layer Stackelberg game approach

Resource management in fog computing is a highly challenging issue. Resource constraints of fog devices, the dynamic and diverse nature of workloads such as data-intensive, computationally intensive, and latency-sensitive applications, and the unpredictability of fog computing environments make it more challenging. How to collaboratively schedule edge, and end multi-layer ubiquitous computing resources to break through single node scale and resource bottlenecks to further improve computing performance also needs to be addressed. In this paper, we leverage idle resources on the user side, and study the resource allocation problems in the D2D-assisted fog computing networks. Firstly, we use an improved Louvain algorithm to cluster users and select the D2D device with the highest priority in each cluster as the cluster head to assist fog computing, where cluster heads can act as both servers and relays. Secondly, considering that fog nodes and D2D devices have limited resources and are not obligated to share their resources for free, we propose a price incentive method to construct a three-layer Stackelberg game model that maximizes the benefits of fog nodes and cluster heads while minimizing user costs. Then, we use the Lagrange multiplier method and the Karush–Kuhn–Tucker (KKT) conditions to solve the equilibrium point of the model to achieve our goal. Finally, the simulation results show that our proposed scheme markedly decreases the overall system cost in comparison to the four benchmark schemes. Moreover, in various scenarios the total cost of our scheme consistently remains lower than the other four schemes, affirming the efficiency of the scheme proposed in this paper.

User Clustering with Spatial Concept using Supervised Learning for NOMA Downlink

This research aims to optimize the performance of Successive Interference Cancellation (SIC) in Power Domain Non-Orthogonal Multiple Access (PD-NOMA) technology by applying spatial concepts through the use of beamforming techniques. User clustering is a key element in achieving this goal, and this research applies various supervised machine learning classification algorithms including Decision Tree, K-Nearest Neighbors (K-NN), Support Vector Machine (SVM), Random Forest, Logistic Regression, and Naive Bayes. The experimental results show that Random Forest achieves the highest accuracy in classifying users, followed by Decision Tree. In addition, in measuring performance using ROC (Receiver Operating Characteristic) and AUC (Area Under the Curve) curves, Decision Tree and Random Forest achieved the best results as well. While in terms of experimentation process time, decision tree has a faster time than random forest. Overall, Random Forest and Decision Tree algorithms are suitable for the task of user clustering in the context of PD-NOMA which utilizes the spatial concept of user to Base Station (BS).

Cluster-Based Strategy for Maximizing the Sum-Rate of a Distributed Reconfigurable Intelligent Surface (RIS)-Assisted Coordinated Multi-Point Non-Orthogonal Multiple-Access (CoMP-NOMA) System.

This article proposes a distributed intelligent Coordinated Multi-Point Non-Orthogonal Multiple-Access (CoMP-NOMA) collaborative transmission model with the assistance of reconfigurable intelligent surfaces (RISs) to address the issues of poor communication quality, low fairness, and high system power consumption for edge users in multi-cellular networks. By analyzing the interaction mechanisms and influencing factors among RIS signal enhancement, NOMA user scheduling, and multi-point collaborative transmission, the model establishes RIS-enhanced edge user grouping and coordinates NOMA user clusters based on this. In the multi-cell RIS-assisted JT-CoMP NOMA downlink transmission, joint optimization of the power allocation (PA), user clustering (UC), and RIS phase-shift matrix design (PS) poses a challenging Mixed-Integer Non-Linear Programming (MINLP) problem. The original problem is decomposed by optimizing the formulas into joint sub-problems of PA, UC, and PA and PS, and solved using an alternating optimization approach. Simulation results demonstrate that the proposed scheme effectively reduces the system's power consumption while significantly improving the system's throughput and rates.

Augmented degree correction for bipartite networks with applications to recommender systems

In recommender systems, users rate items, and are subsequently served other product recommendations based on these ratings. Even though users usually rate a tiny percentage of the available items, the system tries to estimate unobserved preferences by finding similarities across users and across items. In this work, we treat the observed ratings data as partially observed, dense, weighted, bipartite networks. For a class of systems without outside information, we adapt an approach developed for dense, weighted networks to account for unobserved edges and the bipartite nature of the problem. The approach begins with clustering both users and items into communities, and locally estimates the patterns of ratings within each subnetwork induced by restricting attention to one community of users and one community of items community. The local fitting procedure relies on estimating local sociability parameters for every user and item, and selecting the function that determines the degree correction contours which best models the underlying data. We compare the performance of our proposed approach to existing methods on a simulated data set, as well as on a data set of joke ratings, examining model performance in both cases at differing levels of sparsity. On the joke ratings data set, our proposed model performs better than existing alternatives in relatively sparse settings, though other approaches achieve better results when more data is available. Collectively, the results indicate that despite struggling to pick up subtler signals, the proposed approach’s recovery of large scale, coarse patterns may still be useful in practical settings where high sparsity is typical.

Adaptive multibeam hopping in geo satellite networks with non-uniformly distributed ground users

This paper designs a novel low-complexity user-cluster grouping algorithm for adaptive beam hopping in geostationary satellite networks equipped with multibeam phased-array antennas. Each beam serves a cluster of users, and the challenge is to design a beam-hopping pattern where no beams are simultaneously serving nearby user clusters. We develop a line search procedure to identify near-optimum groupings for heterogeneous traffic demands. We provide a necessary condition to determine the boundaries of the line search space. Our approach employs exclusion regions around critical user clusters in congested areas, iterates a sequential congestion-based grouping algorithm, and applies a group-member-swapping procedure. It provides max-min fairness for ground users. Extensive numerical studies have shown that our user grouping algorithm produces near-optimum beam-hopping schedules with low outage probability. It achieves an improvement of up to 13dB in the worst-case signal-to-interference and noise ratio, and doubles the zero-outage data rate, compared to benchmark approaches.

Learning Personalizable Clustered Embedding for Recommender Systems

In recent years, recommender systems have advanced rapidly, where embedding learning for users and items plays a critical role. A standard method learns a unique embedding vector for each user and item. However, such a method has two important limitations in real-world applications: 1) it is hard to learn embeddings that generalize well for users and items with rare interactions; and 2) it may incur unbearably high memory costs when the number of users and items scales up. Existing approaches either can only address one of the limitations or have flawed overall performances. In this paper, we propose Clustered Embedding Learning (CEL) as an integrated solution to these two problems. CEL is a plug-and-play embedding learning framework that can be combined with any differentiable feature interaction model. It is capable of achieving improved performance, especially for cold users and items, with reduced memory cost. CEL enables automatic and dynamic clustering of users and items in a top-down fashion, where clustered entities could jointly learn a shared embedding. The accelerated version of CEL has an optimal time complexity, which supports efficient online updates. Theoretically, we prove the identifiability and the existence of a unique optimal number of clusters for CEL in the context of nonnegative matrix factorization. Empirically, we validate the effectiveness of CEL on three public datasets and one business dataset, showing its consistently superior performance against current state-of-the-art methods. In particular, when incorporating CEL into the business model, it brings an improvement of \(+0.6\% \) in AUC, which translates into a significant revenue gain; meanwhile, the size of the embedding table gets 2650 times smaller. Additionally, we demonstrate that, if there is enough memory, learning a personalized embedding for each user and item around their clustering centers is feasible and can further boost the performance.

The Role of Scientific Research in Human Papillomavirus Vaccine Discussions on Twitter: Social Network Analysis.

Attitudes toward the human papillomavirus (HPV) vaccine and accuracy of information shared about this topic in web-based settings vary widely. As real-time, global exposure to web-based discourse about HPV immunization shapes the attitudes of people toward vaccination, the spread of misinformation and misrepresentation of scientific knowledge contribute to vaccine hesitancy. In this study, we aimed to better understand the type and quality of scientific research shared on Twitter (recently rebranded as X) by vaccine-hesitant and vaccine-confident communities. To analyze the use of scientific research on social media, we collected tweets and retweets using a list of keywords associated with HPV and HPV vaccines using the Academic Research Product Track application programming interface from January 2019 to May 2021. From this data set, we identified tweets referring to or sharing scientific literature through a Boolean search for any tweets with embedded links, hashtags, or keywords associated with scientific papers. First, we used social network analysis to build a retweet or reply network to identify the clusters of users belonging to either the vaccine-confident or vaccine-hesitant communities. Second, we thematically assessed all shared papers based on typology of evidence. Finally, we compared the quality of research evidence and bibliometrics between the shared papers in the vaccine-confident and vaccine-hesitant communities. We extracted 250 unique scientific papers (including peer-reviewed papers, preprints, and gray literature) from approximately 1 million English-language tweets. Social network maps were generated for the vaccine-confident and vaccine-hesitant communities sharing scientific research on Twitter. Vaccine-hesitant communities share fewer scientific papers; yet, these are more broadly disseminated despite being published in less prestigious journals compared to those shared by the vaccine-confident community. Vaccine-hesitant communities have adopted communication tools traditionally wielded by health promotion communities. Vaccine-confident communities would benefit from a more cohesive communication strategy to communicate their messages more widely and effectively.

Unraveling the relationship between coastal landscapes and sentiments: An integrated approach based on social media data and interpretable machine learning methods

AbstractCoastal landscapes exert a significant impact on the human sentimental perceptions and physical and mental well‐being of people. However, little is known about explicitly linking between the landscape characteristics and people's sentimental preferences expressed in social media data. The main objective of this study was to explore the nonlinear and interaction effects of key factors that influenced sentiments in the coastal areas of Hong Kong, considering both subjective landscape preferences and objective landscape patterns. We quantified users' sentiment polarity based on the crowdsourcing textual data of Flickr. To study users' subjective landscape preferences, we computed various visual landscape objects' proportion in images. Meanwhile, eight user clusters and nine image clusters were detected by the identified visual object labels. We quantified objective landscape patterns considering the land use pattens and the availability of public service facilities. Finally, we utilized an interpretable classification model to analyze the factors that may affect sentiments and their interplay interactions. We found that ecotourism‐related clusters exhibited the most positive sentiment. The proportion of floor and sky pixels in images exhibits the highest global relative importance when predicting sentiments. This study extends a new insight on the relationship between landscape characteristics and sentiments from both subjective and objective perspectives based on social media data and interpretable machine learning methods. This research may help decision‐makers in designing landscapes that aptly satisfy to the needs of the public and promote sustainable management of the coastal environment.

Interactive AI Virtual Teaching Resource Intelligent Recommendation Algorithm Based on Similarity Measurement on the Internet of Things Platform

Abstract To tackle the issues of low recommender system correctness and the long time of recommendation in the traditional intelligent recommendation algorithm of interactive artificial intelligence (AI) virtual teaching resources, a similarity measurement-based intelligent recommendation algorithm of interactive AI virtual teaching resources is proposed. According to the contact frequency of adjacent learners, the behavior patterns of learners are mined, and the user characteristics are selected on the basis of the mutual information feature selection method. In the context of the category attribute matrix and learner category attribute scoring matrix of interactive AI virtual teaching resources, the user interest matrix is constructed by the Relevance Feedback based on Inverse Learning Function (RF-ILF) method. The user similarity is calculated and the neighbor set is found in the similar user clusters after clustering. According to the results of the simulation, the suggested algorithm is more efficient and provides more accurate recommendations when recommending interactive AI virtual teaching resources.

Entropy-based detection of Twitter echo chambers.

Echo chambers, i.e. clusters of users exposed to news and opinions in line with their previous beliefs, were observed in many online debates on social platforms. We propose a completely unbiased entropy-based method for detecting echo chambers. The method is completely agnostic to the nature of the data. In the Italian Twitter debate about the Covid-19 vaccination, we find a limited presence of users in echo chambers (about 0.35% of all users). Nevertheless, their impact on the formation of a common discourse is strong, as users in echo chambers are responsible for nearly a third of the retweets in the original dataset. Moreover, in the case study observed, echo chambers appear to be a receptacle for disinformative content.

Graphormer based contrastive learning for recommendation

In recent years, Graph Neural Networks (GNNs) have become a powerful tool for graph representation in recommendation systems. Graphormer integrates its graph structure into a Transformer to provide state-of-the-art performance for graph prediction tasks. Meanwhile, self-supervised contrastive learning has succeeded in processing highly sparse data. Despite these successes, most graph contrastive methods perform random enhancement, such as node/edge perturbation, on a user-item interaction graph, or perform heuristic enhancement techniques, such as user clustering, to generate a new contrastive view. These methods cannot preserve the inherent semantic structure well and are susceptible to noise. In this paper, we propose a Graphormer-based graph contrastive learning method, GO-GCL, which reduces the influence of noise while retaining the inherent semantic structure to some extent, and improves the versatility and robustness of the recommendation based on contrastive learning. The model is represented by Graphormer and then uses singular value decomposition for contrastive enhancement. The experiments on Yelp datasets show that our model has a significant improvement over existing GCL-based approaches. In-depth analyses show the superiority and robustness of GO-GCL in data sparsity and prevalence bias.

Energy-efficient UAV communication: A NOMA scheme with resource allocation and trajectory optimization.

This work investigates a downlink nonorthogonal multiple access (NOMA) scheme with unmanned aerial vehicle (UAV) aided wireless communication, where a single UAV was regarded as an air base station (ABS) to communicate with multiple ground users. Considering the constraints of velocity and maneuverability, a UAV energy efficiency (EE) model was proposed via collaborative design resource allocation and trajectory optimization. Based on this, an EE maximization problem was formulated to jointly optimize the transmit power of ground users and the trajectory of the UAV. To obtain the optimal solutions, this nonconvex problem was transformed into an equivalent convex optimization problem on the basis of three user clustering algorithms. After several alternating iterations, our proposed algorithms converged quickly. The simulation results show an enhancement in EE with NOMA because our proposed algorithm is nearly 99.6% superior to other OMA schemes.

Optimization of Product Placement on E-commerce Platforms with K-Means Clustering to Improve User Experience

This study delves into product placement strategies on E-commerce platforms using K-Means Clustering analysis. Employing an experimental methodology, data about products and user preferences were gathered to delineate product and user clusters. The K-Means Clustering analysis yielded three primary product clusters and four user preference clusters. These findings hold significant practical implications, empowering E-commerce platforms to refine user experience personalization, streamline sales efficiency, and bolster overall business performance. Platforms can positively influence sales conversion rates and user satisfaction by implementing targeted and adaptable product placement strategies. This research contributes not only to the theoretical comprehension of product placement in E-commerce but also furnishes actionable insights for stakeholders to optimize platform operations and deliver an enriched online shopping experience.

Uplink-downlink resource optimization for provisioning 5G application considering SUI fading channel models

Recently, superposition coding (SC), Stanford University Interim (SUI) fading channel, user clustering, successive interference cancellation (SIC) mechanism has been incorporated into the power allocation design for attaining good performance. Nonetheless, the existing model works with cluster size up to 4 user per cluster and induce higher control channel overhead. In addressing this paper introduce an effective resource allocation (ERA) design for power domain non-orthogonal multiple access (PD-NOMA) system. The ERA, first employ Zadoff Chu (ZC) coding, then power is allocated according to the distance considering 5 user per cluster. Finally, multi-level SIC is done at the receiver end. The ERA performance is studied by varying signal-to-noise ratio (SNR) in terms of bit error rate (BER) considering a maximum of five users per cluster. Further, the ERA performance in terms of BER is tested under different scenario defined in SUI propagation scenario such as higher pathloss, moderate path loss, and low path loss. The throughput performance under varied density and speed is also studied; the ERA achieves much better performance in terms of ERA, SNR and throughput than standard multi-user downlink resource allocation (MUDRA) model.

A taxonomy on power optimization techniques for fifth-generation heterogenous non-orthogonal multiple access networks

Non-orthogonal multiple access (NOMA) is an anticipated technology for fifth-generation networks for increasing mass connectivity, spectrum efficiency, user-fairness, and higher capacity. NOMA allows end-clients to share indistinguishable radio resources such as spreading code, subcarrier, and time slots simultaneously. Thus, the main challenge involved in conceptualizing effective NOMA design is selection of resource allocation (i.e., user clustering, power allocation, and quality-of-service (QoS) assurance) algorithms. NOMA can be easily integrated with current fifth-generation multi-access methodologies. In this survey paper, the NOMA methodologies are discussed, and provide an overview of the methodologies and algorithms designed for optimizing power allocation, interference management, and network selection management in the heterogenous multiple carrier NOMA. The survey highlights the current limitation of the existing resource provisioning framework and presents a solution to overcome the current limitation.

Quantum aided efficient resource control for connected support in IRS assisted networks

Context:To achieve the vision of all connected world with uninterrupted communication support, 6G technology plays an important role. But the scarce radio spectrum and limited network resources is the main challenge in delivering its promised performance. Objectives:This paper presents an IRS-aided cell free NOMA network model that aims to provide uniform network coverage. The future 6G technology envisions for serving billions of interconnected devices with seamless communication support, data handling capabilities and computational accuracy. But the scarcity of network resources is the main limitation. Thus, the need is to design quantum enabled intelligent and dynamic networks capable of offering extended network capabilities. Proposed work is on intelligent network framework that provides uniform network coverage through efficient resource management for a 6G enabled expanded IoT network. Methods:To enable efficient resource management, a quantum enabled resource control algorithm is proposed that creates user clusters and associates each AP-IRS pair to each cluster. Each AP transmits the superimposed signals of its intended cluster against all the user clusters as in conventional NOMA system. The nodes in each cluster have been assigned unique pilots so as to avoid intracluster interference. The use of IRS enables desired NOMA beamforming such that the effect of unfavourable wireless environment is mitigated. Results:The performance of the IRS-aided cell-free NOMA network is evaluated for average sum rate with different AP transmit power, cluster sizes, IRS reflecting elements and IRS phase shifts. It is shown that at transmit power per AP of 30dBm, the average sum rate of the system improves by 7.52% with the proposed algorithm using equal power allocation scheme. Further, the comparative performance analysis of three different communication systems is carried out to validate the proposed communication model. Conclusion:It is observed that with more number of users per cluster, the average sum rate of the system initially increases for small cluster sizes and then it becomes constant for large cluster sizes. The proposed clustering method outperforms the random clustering approach achieving sum rate of 12.9 bits/s/Hz with N =300 and M= 8. The comparison of different communication scenarios reveals that the maximum sum rate of 12.2 bits/s/Hz is achieved with the proposed model incorporating proposed clustering mechanism. Further, the energy efficiency analysis suggests that energy efficiency improves with N and Pc with proposed clustering approach. The use case scenarios for the integration of quantum computing with IRS technology are also presented.

Federated Contextual Cascading Bandits with Asynchronous Communication and Heterogeneous Users

We study the problem of federated contextual combinatorial cascading bandits, where agents collaborate under the coordination of a central server to provide tailored recommendations to users. Existing works consider either a synchronous framework, necessitating full agent participation and global synchronization, or assume user homogeneity with identical behaviors. We overcome these limitations by considering (1) federated agents operating in an asynchronous communication paradigm, where no mandatory synchronization is required and all agents communicate independently with the server, (2) heterogeneous user behaviors, where users can be stratified into latent user clusters, each exhibiting distinct preferences. For this setting, we propose a UCB-type algorithm with delicate communication protocols. Through theoretical analysis, we give sub-linear regret bounds on par with those achieved in the synchronous framework, while incurring only logarithmic communication costs. Empirical evaluation on synthetic and real-world datasets validates our algorithm's superior performance in terms of regrets and communication costs.

Current and potential use of animal disease data by stakeholders in the global south and north

What cannot be measured will not be managed. The Global Burden of Animal Diseases (GBADs) will generate information on animal disease burdens by species, production system, type and gender of farmer and consumer, geographical region, and time period. To understand the demand for burden of animal disease (BAD) data and how end-users might benefit from this, we reviewed the literature on animal diseases prioritisation processes (ADPP) and conducted a survey of BAD information users. The survey covered their current use of data and prioritizations as well as their needs for different, more, and better information. We identified representative (geography, sector, species) BAD experts from the authors’ networks and publicly available documents and e-mailed 1485 experts. Of 791 experts successfully contacted, 271 responded (34% response rate), and 185 complete and valid responses were obtained. Most respondents came from the public sector followed by academia/research, and most were affiliated to institutions in low- and middle-income countries (LMICs). Of the six ADPPs commonly featured in literature, only three were recognised by more than 40% of experts. An additional 23 ADPPs were used. Awareness of ADDPs varied significantly by respondents. Respondents ranked animal disease priorities. We used exploded logit to combine first, second and third disease priorities to better understand prioritzation and their determinants. Expert priorities differed significantly from priorities identified by the ADDPs, and also from the priorities stated veterinary services as reported in a survey for a World Organisation of Animal Health (WOAH) technical item. Respondents identified 15 different uses of BAD data. The most common use was presenting evidence (publications, official reports, followed by disease management, policy development and proposal writing). Few used disease data for prioritzation or resource allocation, fewer routinely used economic data for decision making, and less than half were aware of the use of decision support tools (DSTs). Nearly all respondents considered current BAD metrics inadequate, most considered animal health information insufficiently available and not evidence-based, and most expressed concerns that decision-making processes related to animal health lacked transparency and fairness. Cluster analysis suggested three clusters of BAD users and will inform DSTs to help them better meet their specific objectives. We conclude that there is a lack of satisfaction with current BAD information, and with existing ADDPs, contributing to sub-optimal decision making. Improved BAD data would have multiple uses by different stakeholders leading to better evidenced decisions and policies; moreover, clients will need support (including DSTs) to optimally use BAD information.

Evolution of the profiles of new psychotropic drug users before and during the COVID-19 crisis: an original longitudinal approach through multichannel sequence analysis using the French health-care database.

The COVID-19 pandemic has had a substantial impact on mental health. An increase in the use of anxiolytic, hypnotic, and antidepressant drugs has been highlighted in France, but with no information at the individual level (trajectories) or concerning patient characteristics. The objective of this study was to describe the profile of new psychotropic drug users since the beginning of the pandemic. We formed two historical cohorts using the Pays-de-la-Loire regional component of the National Health Data System (SNDS): a "COVID-19 crisis cohort" (2020-2021) and a "control cohort" (2018-2019). We analyzed reimbursements for psychotropic medications (anxiolytics, antidepressants, hypnotics, mood stabilizers, and antipsychotics) using a multichannel sequence analysis and performed clustering analysis of sequences. The proportion of new consumers of psychotropic drugs was higher in the COVID-19 crisis cohort (18.0%) than that in the control cohort (16.0%). In the COVID-19 cohort, three clusters of psychotropic drug users were identified, whereas four clusters were identified in the control cohort. A time lag in treatment initiation was observed in the COVID-19 crisis cohort (September) compared with the control cohort (July). This study is one of the first to analyze the profile of psychotropic treatment users during the COVID-19 crisis. Our analysis sheds light on changes in patterns of psychotropic drug use during the COVID-19 pandemic, possibly associated with changes in prescribing conditions and mental health conditions during the crisis. This study also provides an example of the application of an innovative longitudinal analysis methodology in the field of pharmacoepidemiology.