Preference Graph Research Articles

Adaptive multi-graph contrastive learning for bundle recommendation

Recently, recommending bundles - sets of items that complement each other - instead of individual items to users has drawn much attention in both academia and industry. Models based on Graph Neural Networks (GNNs) for bundle recommendation have achieved great success in capturing users’ preferences by modeling pairwise correlations among users, bundles, and items via information propagation on graphs. However, a notable limitation lies in their insufficient focus on explicitly modeling intricate ternary relationships. Additionally, the loose combination of node embeddings from different graphs tends to introduce noise, as it fails to consider disparities among the graphs. To this end, we propose a novel approach called Adaptive Multi-Graph Contrastive Learning for Bundle Recommendation (AMCBR). Specifically, AMCBR models ternary interactions by constructing multiple graphs, including a bundle preference graph based on direct user-bundle interactions, a collaborative neighborhoods graph featuring user-level and bundle-level subgraphs, and an item-level preference hypergraph capturing indirect user-bundle relationships through items. Then, (hyper)graph convolution is applied to each (hyper)graph to encode diverse potential preferences into node embeddings. To enhance the model’s robustness, an adaptive aggregation module is employed to assign varying weights to node embeddings from different graphs during the fusion process, which enriches the semantic and comprehensive information in the embeddings while mitigating potential noise. Finally, a contrastive learning strategy is proposed to jointly optimize the model, strengthening collaborative links between individual graphs. Extensive experiments on three real datasets demonstrate that AMCBR can outperform the state-of-the-art baselines on the Top-K recommendations.

Intelligent maritime question-answering and recommendation system based on maritime vessel activity knowledge graph

Traditional maritime traffic management typically relies on positioning data for data mining without incorporating other multi-source data to analyze the maritime vessel activity, which cannot conduct comprehensive maritime knowledge mining. Thus, this study integrates multi-source data, such as trajectory, maritime accident text, and geographic data, to create a maritime vessel activity knowledge graph. On this basis, a question-answering model is developed based on a bidirectional question-answering attention graph neural network, and a personalized recommendation model is developed based on an attention-enhanced joint knowledge propagation and a user preference graph neural network. The former assists users in extracting valuable information from the maritime vessel activity knowledge graph, while the latter predicts the users' potential interests and automatically recommends vessel entities based on their historical query information. Experimental results show that the proposed question-answering model improved the F1-score by 2.31%–10.09% compared to state-of-the-art baseline models on the MVA question-answering dataset. Similarly, the proposed personalized recommendation model improved the click-through rate prediction accuracy by 2.46%–7.05% compared to state-of-the-art baseline models on the MVA personalized recommendation dataset.

Allocation of indivisible items with individual preference graphs

Allocation of indivisible items with individual preference graphs

Relation-propagation meta-learning on an explicit preference graph for cold-start recommendation

The cold-start problem has been of great concern in the recommendation domain. To address this problem, meta-learning frameworks have been widely adopted due to their fast adaptation to new tasks. However, existing meta-learning-based methods always use higher-order graph structures to obtain global user preferences but neglect to consider local preferences at different rating levels in a fine-grained manner. Since differences in user ratings on items truly reflect differences in users’ local preferences, we propose a relation-propagation meta-learning on explicit preference graph for cold-start recommendation (RPMLG-Rec) to improve the generalization performance of existing meta-learning-based methods. Specifically, we focus on capturing the relationships between local preferences. First, our RPMLG-Rec approach concatenates different local preferences to form the nodes of local user preference and further constructs an explicit preference graph. Second, the relationships between local preferences, including intraclass commonality and interclass uniqueness, are used to guide the propagation of relationships in the explicit preference graph with graph convolutional networks and produce more distinguishable local preference nodes. Third, the precise global user preference is obtained with an attention mechanism. Finally, prior knowledge is learned based on a set of training tasks and quickly adapted to make recommendations for new tasks, following the optimization-based meta-learning training strategy. To the best of our knowledge, this is the first time that the relationships between local user preference nodes have been explicitly considered in cold-start recommendation. In addition, we conducted extensive experiments on two real-world datasets, and the experimental results demonstrate the effectiveness of our proposed framework.

Three-way Preference Completion via Preference Graph

With the personal partial rankings from agents over a subset of alternatives, the goal of preference completion is to infer the agent’s personalized preference over all alternatives including those the agent has not yet handled from uncertain preference of third parties. By combining the partial rankings of the target agent and the partial rankings from third parties to settle some disagreement with three-way preference completion, which includes a general strategy, an optimal strategy, and a pessimistic strategy, it forms the weighted preference graph. Technically, to settle the disagreement and obtain the completed preference of the target agent in the weighted preference graph, maximum likelihood estimation (MLE) under Mallows is proposed and validated theoretically by removing edges with the minimum weight in the weighted preference graph. However, it is not easy to locate the edges with the minimum weight efficiently in a big graph. Hence, an optimal MLE algorithm and three greedy MLE algorithms are proposed to process the MLE. Furthermore, these proposed algorithms are experimentally validated and compared with each other by both the synthetic dataset and the Flixter dataset.

Learning the User’s Deeper Preferences for Multi-modal Recommendation Systems

Recommendation system plays an important role in the rapid development of micro-video sharing platform. Micro-video has rich modal features, such as visual, audio, and text. It is of great significance to carry out personalized recommendation by integrating multi-modal features. However, most of the current multi-modal recommendation systems can only enrich the feature representation on the item side, while it leads to poor learning of user preferences. To solve this problem, we propose a novel module named Learning the User’s Deeper Preferences (LUDP) , which constructs the item-item modal similarity graph and user preference graph in each modality to explore the learning of item and user representation. Specifically, we construct item-item similar modalities graph using multi-modal features, the item ID embedding is propagated and aggregated on the graph to learn the latent structural information of items; The user preference graph is constructed through the historical interaction between the user and item, on which the multi-modal features are aggregated as the user’s preference for the modal. Finally, combining the two parts as auxiliary information enhances the user and item representation learned from the collaborative signals to learn deeper user preferences. Through a large number of experiments on two public datasets (TikTok, Movielens), our model is proved to be superior to the most advanced multi-modal recommendation methods.

Fair subgraph selection for contagion containment (Brief Announcement)

We present a new class of problems where the goal is to select a “fair” subgraph H of a given graph G = (V,E), such that H decomposes into many small components. A subgraph Hc G is (P,d) fair if every vertex vϵ P has the same degree d in H, where PcV and d > 0 are input parameters.These problems arise when the goal is to allow individuals to equally participate in activities in such a way that the connected components within an interaction graph, which models potential interactions among people, are of the smallest possible size, so that the spread of the contagion, and the difficulty of contact tracing in case of infection, is minimized. Within a preference graph that models the set of preferred choices for each individual when selecting among available options of where to conduct any particular type of activity (e.g., which gym to attend), we seek to compute the fair subgraph of assignments of individuals to these options, so that the number of people in each connected component (“interaction community”) of the resulting subgraph is minimized, and everyone is given the same number of options for every activity.We show that the fair subgraph selection problem is NP-hard, even for very restricted versions. We then formulate the problem as an integer program, and give a polynomial time computable lower bound on the optimal solution.

FairSR: Fairness-aware Sequential Recommendation through Multi-Task Learning with Preference Graph Embeddings

Sequential recommendation (SR) learns from the temporal dynamics of user-item interactions to predict the next ones. Fairness-aware recommendation mitigates a variety of algorithmic biases in the learning of user preferences. This article aims at bringing a marriage between SR and algorithmic fairness. We propose a novel fairness-aware sequential recommendation task, in which a new metric, interaction fairness , is defined to estimate how recommended items are fairly interacted by users with different protected attribute groups. We propose a multi-task learning-based deep end-to-end model, FairSR, which consists of two parts. One is to learn and distill personalized sequential features from the given user and her item sequence for SR. The other is fairness-aware preference graph embedding (FPGE). The aim of FPGE is two-fold: incorporating the knowledge of users’ and items’ attributes and their correlation into entity representations, and alleviating the unfair distributions of user attributes on items. Extensive experiments conducted on three datasets show FairSR can outperform state-of-the-art SR models in recommendation performance. In addition, the recommended items by FairSR also exhibit promising interaction fairness.

A Method to Minimize the Effort for Damper–Blade Matching Demonstrated on Two Blade Sizes

A method called PCR (Platform Centered Reduction) is designed to more effectively perform complex iterative and nonlinear calculations required for the dynamic response of turbine blades damped by dry friction contacts between rigid dampers and airfoil-to-neck platform. The key feature of PCR is to represent all nonlinear forces on the blade platform by means of only six degrees of freedom at a point located within the platform volume, regardless of the number of damper–platform contact elements. Despite reducing the effort and computational time by more than one order of magnitude, the method proves to be fully accurate by a check against the corresponding nonlinear Finite Elements (FE) calculation. It is also shown that the limit exciting force, indicating the upper capability to dampen vibrations, can be calculated with a simple linear modal analysis. In order to search for the best blade–damper match, the preferred graph represents relevant bending stresses on the airfoil against excitation forces. A detailed application of the method concerns two significantly different blade sizes, by varying parameters such as neck length and damper centrifugal force. Finally, it is emphasized that a final check by a complete FE analysis is still possible as a purely linear solution fed by sets of contact forces previously determined through the PCR at any desired frequency and excitation.

Product Module Attribute Parameter Configuration Model considering Customer Requirements Preferences

In the process of modular product configuration, it is necessary to transform customer requirements into product module attributes (PMA) parameters. However, previous research lacks consideration about customer requirement preference in the process of this transformation. First, we use a preference graph (PG) to obtain the customer preference weight vector for the requirement node. Second, on the basis of traditional Quality Function Deployment (QFD), the method of fuzzy correlation evaluation is introduced to get the correlation value between module attributes, and the combination programming model of PMA is further obtained by synthesizing the preference weight vector. Finally, the final configuration scheme is obtained by solving the model with the genetic algorithm. By integrating the weights of the above-mentioned nodes, the similarity of the product case is obtained, and a more satisfied case of the customer is obtained. Taking the automated guided vehicle car product as an example, the effectiveness and practicability of the proposed method are verified.

Relationships between the response of the sweet taste receptor, salivation toward sweeteners, and sweetness intensity.

Sweeteners are widely used in food products, and their sweetness potency is usually evaluated by comparing it with that of sucrose. This, however, has led to confusion as some sweeteners are evaluated based on their maximum value of sweet taste response, while others are evaluated by their threshold value. Here, we aimed to develop a novel nonverbal sweetness evaluation system through the sweet taste signal transduction by comparing the responses of the sweet taste receptor, salivation, taste intensity, and preference among six sweeteners. The hT1r2/hT1r3 sweet taste receptor responses represented the input signal of the sweet taste signal transduction, while salivation, sweet taste intensity, and participants' preferences represented the output signals by the gustatory–salivary reflex, primary gustatory cortex area, and the secondary gustatory cortex, respectively. Our results showed that the sweet taste receptor, sweet intensity, and salivary secretion responses were concentration‐dependent and expressed exponentially. Moreover, the results comparing coefficients showed 15–35 times more sensitivity between the response of hT1r2/hT1r3 and the salivation or the sweet taste intensity in non‐nutrient sweeteners. The preference graph curve was not exponential, suggesting that the sweetener preference was not related to the sweet taste receptor, salivation, or sweet taste intensity. These results may suggest that the sweet taste signal of the non‐nutritive sweeteners might be maintained by taste reception by hT1r2/hT1r3 to taste recognition in the primary gustatory area and that receptor responses and salivation could be used as indicators of sweetness intensity.

Modeling design information systems with multi-criteria

This work is devoted to the development of a new method for modeling information systems of multi-criteria design, based on expert ranking of criteria. The possibilities of the fuzzy-set theory and the possibility theory allow for a comprehensive analysis of criteria by reducing numerical indicators to a qualitative form, passing to the corresponding fuzzy relations of preference. Some of the information may be lost, therefore conclusions should be applied, both on the basis of “quantitative” processing of the results of measuring the criteria, and on the basis of “qualitative” processing. If these conclusions coincide, then their adequacy will be confirmed based on the initial data. Since the modeling of complex design systems is determined by several criteria defined on different base sets, and the problem of determining the multi-criteria efficiency in conditions of incompleteness of the initial data remains relevant, within the framework of this article, an optimization method is proposed simultaneously according to many criteria with incompleteness of the initial data. The difference between the described method lies in the fact that the expert ranking of criteria enables to formally set fuzzy preference graphs that make it possible to obtain comprehensive information about estimates in the rank scale during pairwise comparisons, with the aim of subsequent search for Pareto-optimal solutions in multi-criteria problems.

An Analysis Framework for Metric Voting based on LP Duality

Distortion-based analysis has established itself as a fruitful framework for comparing voting mechanisms. m voters and n candidates are jointly embedded in an (unknown) metric space, and the voters submit rankings of candidates by non-decreasing distance from themselves. Based on the submitted rankings, the social choice rule chooses a winning candidate; the quality of the winner is the sum of the (unknown) distances to the voters. The rule's choice will in general be suboptimal, and the worst-case ratio between the cost of its chosen candidate and the optimal candidate is called the rule's distortion. It was shown in prior work that every deterministic rule has distortion at least 3, while the Copeland rule and related rules guarantee distortion at most 5; a very recent result gave a rule with distortion 2 + √5 ≈ 4.236.We provide a framework based on LP-duality and flow interpretations of the dual which provides a simpler and more unified way for proving upper bounds on the distortion of social choice rules. We illustrate the utility of this approach with three examples. First, we show that the Ranked Pairs and Schulze rules have distortion Θ(√n). Second, we give a fairly simple proof of a strong generalization of the upper bound of 5 on the distortion of Copeland, to social choice rules with short paths from the winning candidate to the optimal candidate in generalized weak preference graphs. A special case of this result recovers the recent 2 + √5 guarantee. Finally, our framework naturally suggests a combinatorial rule that is a strong candidate for achieving distortion 3, which had also been proposed in recent work. We prove that the distortion bound of 3 would follow from any of three combinatorial conjectures we formulate.

Preference Graph of Potential Method as a Fuzzy Graph

An autocatalytic set (ACS) is a graph. On the other hand, the Potential Method (PM) is an established graph based concept for optimization purpose. Firstly, a restricted form of ACS, namely, weak autocatalytic set (WACS), a derivation of transitive tournament, is introduced in this study. Then, a new mathematical concept, namely, fuzzy weak autocatalytic set (FWACS), is defined and its relations to transitive PM are established. Some theorems are proven to highlight their relations. Finally, this paper concludes that any preference graph is a fuzzy graph Type 5.

Research on solving method of similar cases considering customer preference

Customer requirement preference is an important part of customer satisfaction. In view of similar case retrieval technology for existing product level, in the process of solving similar cases, there is no consideration for customer requirement preference. This article proposes a similar case solution method considering customer requirement preference. First, we deal with the expression of customer requirements and transform them into operable parameter forms according to the mapping model. Second, the preference graph is used to analyze the customer’s requirement preference, to determine the preference weight, and to weigh the final weight of the requirement node with the initial weight determined by the fuzzy analytic hierarchy process. Finally, the similarity degree solving model of requirement node and product case attribute parameters is established. By integrating the weights of the above-mentioned nodes, the similarity of the product case is obtained, and a more satisfied case of the customer is obtained. Taking the automated guided vehicle car product as an example, the effectiveness of the proposed method is verified.

Consensus-Based Multi-Person Decision Making Using Consistency Fuzzy Preference Graphs

This paper presents consensus based multi-person decision making (MPDM) using consistency graphs (additive consistent and order consistent) in a fuzzy environment. At the first level, consistency analysis is put forward after defining consistent fuzzy preference graph (CFPG) with the help of additive transitivity. This analysis further leads us to determine priority weights vector of the decision-makers (DMRs) after evaluation consistency weights. At the second level, the consensus analysis helps us to determine whether the selection process should be initiated or not. If the consensus degree amongst DMRs does not reach a minimum acceptable level, then the enhancement mechanism plays a central role to improve the consensus level by giving suitable suggestions to DMRs. In the end, the weighted sum operator (WSO) is used to get aggregated consistency fuzzy preference graph (AgCFPG) and the order consistency property provides us sufficient information to rank the alternatives.

A Comprehensive Algorithm for Evaluating Node Influences in Social Networks Based on Preference Analysis and Random Walk

In the era of big data, social network has become an important reflection of human communications and interactions on the Internet. Identifying the influential spreaders in networks plays a crucial role in various areas, such as disease outbreak, virus propagation, and public opinion controlling. Based on the three basic centrality measures, a comprehensive algorithm named PARW-Rank for evaluating node influences has been proposed by applying preference relation analysis and random walk technique. For each basic measure, the preference relation between every node pair in a network is analyzed to construct the partial preference graph (PPG). Then, the comprehensive preference graph (CPG) is generated by combining the preference relations with respect to three basic measures. Finally, the ranking of nodes is determined by conducting random walk on the CPG. Furthermore, five public social networks are used for comparative analysis. The experimental results show that our PARW-Rank algorithm can achieve the higher precision and better stability than the existing methods with a single centrality measure.

An ordinal consistency-based group decision making process with probabilistic linguistic preference relation

The probabilistic linguistic preference relation (PLPR) reflects different preference degrees (or weights) of experts over possible linguistic terms and improves the flexibility for experts in expressing linguistic preference information. Ordinal consistency is the minimum condition to guarantee that experts would not give some self-contradictory preferences. The ranking result of alternatives based on ordinal inconsistent preference relations is unreliable. This study investigates the ordinal consistency for PLPRs and the identification and modification of ordinal inconsistent PLPRs. To do so, this paper first defines the ordinal consistency of PLPR and presents the preference graph associated with the PLPR. Then, we present theorems about ordinal consistency of PLPR and define an ordinal consistency index to measure the ordinal consistency level from the perspective of preference graph. Moreover, we put forward an algorithm to check whether a PLPR is ordinal consistent and further identify the inconsistent elements for the inconsistent PLPR. Another algorithm based on three modification principles is further developed to modify the ordinal inconsistent PLPR and obtain the ordinal consistent one. Furthermore, we introduce new aggregation operators and present the whole decision-making procedure based on the ordinal consistency. Finally, a case study about water resource management is illustrated. Discussions of results and comparison analysis are also provided.

Communicating projected survival with treatments for chronic kidney disease: patient comprehension and perspectives on visual aids

BackgroundMortality in end stage renal disease (ESRD) is higher than many malignancies. There is no data about the optimal way to present information about projected survival to patients with ESRD. In other areas, graphs have been shown to be more easily understood than narrative. We examined patient comprehension and perspectives on graphs in communicating projected survival in chronic kidney disease (CKD).MethodsOne hundred seventy-seven patients with CKD were shown 4 different graphs presenting post transplantation survival data. Patients were asked to interpret a Kaplan Meier curve, pie chart, histogram and pictograph and answer a multi-choice question to determine understanding.ResultsWe measured interpretation, usefulness and preference for the graphs. Most patients correctly interpreted the graphs. There was asignificant difference in the percentage of correct answers when comparing different graph types (p = 0.0439). The pictograph was correctly interpreted by 81% of participants, the histogram by 79%, pie chart by 77% and Kaplan Meier by 69%. Correct interpretation of the histogram was associated with educational level (p = 0.008) and inversely associated with age > 65 (p = 0.008). Of those who interpreted all four graphs correctly, there was an association with employment (p = 0.001) and New Zealand European ethnicity (p = 0.002).87% of patients found the graphs useful. The pie chart was the most preferred graph (p 0.002).The readability of the graphs may have been improved with an alternative colour choice, especially in the setting of visual impairment.ConclusionVisual aids, can be beneficial adjuncts to discussing survival in CKD.

Operators of preference composition for CP-nets

Abstract Now the handling of user preference is becoming an increasingly important issue in database fields where they capture soft criteria for queries. A broader category of qualitative preferences with dependent relations among multiple attributes is widely existing, which is CP-nets. In this article, we focus on designing the operators of preference composition for CP-nets. Firstly, we extend Pareto composition to our model by including equivalence relation ≈, incomparability relation ∥ and conflicting relation ⊥, which can preserve a strict partial order and conditional associativity. On this basis, two questions are solved: (a) the generation of satisfiability sequences for CP-nets, (b) the top - k queries of relational database with CP-nets preference. For (a), a CP-net is induced into multiple tables, consequently the strong dominance tests between outcomes can be solved by using preference composition instead of using induced preference graph of CP-nets. For (b), we adopt the concept of Query Lattice to provide a natural semantics for the block sequence answering a preference query, where two algorithms (called QOCP and IQOCP ) are introduced. These questions are solved efficiently and effectively at the perspective of combination of graph model and relational database.