Link Prediction Model Research Articles

MCoGCN-motif high-order feature-guided embedding learning framework for social link prediction

Traditional social link prediction models primarily concentrate on the adjacency features of the network, overlooking the rich high-order structural information within. Therefore, the study of effective extraction and encoding of these high-order features, and their integration into prediction models, holds significant theoretical and practical value. To address this challenge, we propose a novel embedding learning framework guided by motif high-order features for social link prediction tasks. Firstly, we utilize the motif adjacency matrix to capture complex patterns in social networks. Through a propagation process, node embeddings can carry the structural information of the network. Subsequently, we design a simplified attention mechanism, allowing embeddings carrying motif high-order features to guide the representation of embeddings based on adjacency features. We then employ a feed-forward neural network to optimize node embeddings. Specifically, this framework addresses the issue of weakly correlated nodes in the network, which struggle to learn effective embeddings due to a lack of direct information. By guiding with high-order motif features, the framework enhances the similarity and predictive power of these node embeddings. Finally, we conducted a detailed evaluation of the predictive performance of our model on four social networks. The experimental results indicate that our model exhibits high accuracy and advantages in predicting social links.

Explicit and Implicit Feature Contrastive Learning Model for Knowledge Graph Link Prediction.

Knowledge graph link prediction is crucial for constructing triples in knowledge graphs, which aim to infer whether there is a relation between the entities. Recently, graph neural networks and contrastive learning have demonstrated superior performance compared with traditional translation-based models; they successfully extracted common features through explicit linking between entities. However, the implicit associations between entities without a linking relationship are ignored, which impedes the model from capturing distant but semantically rich entities. In addition, directly applying contrastive learning based on random node dropout to link prediction tasks, or limiting it to triplet-level, leads to constrained model performance. To address these challenges, we design an implicit feature extraction module that utilizes the clustering characteristics of latent vector space to find entities with potential associations and enrich entity representations by mining similar semantic features from the conceptual level. Meanwhile, the subgraph mechanism is introduced to preserve the structural information of explicitly connected entities. Implicit semantic features and explicit structural features serve as complementary information to provide high-quality self-supervised signals. Experiments are conducted on three benchmark knowledge graph datasets. The results validate that our model outperforms the state-of-the-art baselines in link prediction tasks.

SEGODE: a structure-enhanced graph neural ordinary differential equation network model for temporal link prediction

SEGODE: a structure-enhanced graph neural ordinary differential equation network model for temporal link prediction

Adversarial nonnegative matrix factorization for temporal link prediction

Temporal link prediction has been extensively studied and widely applied in various applications, aiming to predict future network links based on the historical networks. However, most existing methods ignore the behavior of previous network updating information in temporal networks. To address these issues, we propose a novel link prediction model based on adversarial nonnegative matrix factorization, which fuses graph representation and adversarial learning to perform temporal link prediction. Specifically, we add a bounded adversary matrix to the input matrix to provide the robustness against real perturbations. Then, our model fully exploits the impact of snapshots by using communicability. Simultaneously, we utilize the cosine similarity to extract the node similarity and map it to low-dimensional latent representation to preserve the local structure. Additionally, we provide effective updating rules to learn the parameters of this model. Extensive experiments results on six real-world networks demonstrate that the proposed method outperforms several classical and the state-of-art matrix-based methods.

FuseLinker: Leveraging LLM’s pre-trained text embeddings and domain knowledge to enhance GNN-based link prediction on biomedical knowledge graphs

ObjectiveTo develop the FuseLinker, a novel link prediction framework for biomedical knowledge graphs (BKGs), which fully exploits the graph’s structural, textual and domain knowledge information. We evaluated the utility of FuseLinker in the graph-based drug repurposing task through detailed case studies. MethodsFuseLinker leverages fused pre-trained text embedding and domain knowledge embedding to enhance the graph neural network (GNN)-based link prediction model tailored for BKGs. This framework includes three parts: a) obtain text embeddings for BKGs using embedding-visible large language models (LLMs), b) learn the representations of medical ontology as domain knowledge information by employing the Poincaré graph embedding method, and c) fuse these embeddings and further learn the graph structure representations of BKGs by applying a GNN-based link prediction model. We evaluated FuseLinker against traditional knowledge graph embedding models and a conventional GNN-based link prediction model across four public BKG datasets. Additionally, we examined the impact of using different embedding-visible LLMs on FuseLinker’s performance. Finally, we investigated FuseLinker’s ability to generate medical hypotheses through two drug repurposing case studies for Sorafenib and Parkinson’s disease. ResultsBy comparing FuseLinker with baseline models on four BKGs, our method demonstrates superior performance. The Mean Reciprocal Rank (MRR) and Area Under receiver operating characteristic Curve (AUROC) for KEGG50k, Hetionet, SuppKG and ADInt are 0.969 and 0.987, 0.548 and 0.903, 0.739 and 0.928, and 0.831 and 0.890, respectively. ConclusionOur study demonstrates that FuseLinker is an effective novel link prediction framework that integrates multiple graph information and shows significant potential for practical applications in biomedical and clinical tasks. Source code and data are available at https://github.com/YKXia0/FuseLinker.

Structural link prediction model with multi-view text semantic feature extraction

The exponential expansion of information has made text feature extraction based on simple semantic information insufficient for the multidimensional recognition of textual data. In this study, we construct a text semantic structure graph based on various perspectives and introduce weight coefficients and node clustering coefficients of co-occurrence granularity to enhance the link prediction model, in order to comprehensively capture the structural information of the text. Firstly, we jointly build the semantic structure graph based on three proposed perspectives (i.e., scene semantics, text weight, and graph structure), and propose a candidate keyword set in conjunction with an information probability retrieval model. Subsequently, we propose weight coefficients of co-occurrence granularity and node clustering coefficients to improve the link prediction model based on the semantic structure graph, enabling a more comprehensive acquisition of textual structural information. Experimental results demonstrate that our research method can reveal potential correlations and obtain more complete semantic structure information, while the WPAA evaluation index validates the effectiveness of our model.

Context propagation based influence maximization model for dynamic link prediction

Influence maximization (IM) in dynamic social networks is an optimization problem to analyze the changes in social networks for different periods. However, the existing IM methods ignore the context propagation of interaction behaviors among users. Hence, context-based IM in multiplex networks is proposed here. Initially, multiplex networks along with their contextual data are taken as input. Community detection is performed for the network using the Wilcoxon Hypothesized K-Means (WH-KMA) algorithm. From the detected communities, the homogeneous network is used for extracting network topological features, and the heterogeneous networks are used for influence path analysis based on which the node connections are weighted. Then, the influence-path-based features along with contextual features are extracted. These extracted features are given for the link prediction model using the Parametric Probability Theory-based Long Short-Term Memory (PPT-LSTM) model. Finally, from the network graph, the most influencing nodes are identified using the Linear Scaling based Clique (LS-Clique) detection algorithm. The experimental outcomes reveal that the proposed model achieves an enhanced performance.

FoodAtlas: Automated knowledge extraction of food and chemicals from literature

Automated generation of knowledge graphs that accurately capture published information can help with knowledge organization and access, which have the potential to accelerate discovery and innovation. Here, we present an integrated pipeline to construct a large-scale knowledge graph using large language models in an active learning setting. We apply our pipeline to the association of raw food, ingredients, and chemicals, a domain that lacks such knowledge resources. By using an iterative active learning approach of 4120 manually curated premise-hypothesis pairs as training data for ten consecutive cycles, the entailment model extracted 230,848 food-chemical composition relationships from 155,260 scientific papers, with 106,082 (46.0 %) of them never been reported in any published database. To augment the knowledge incorporated in the knowledge graph, we further incorporated information from 5 external databases and ontology sources. We then applied a link prediction model to identify putative food-chemical relationships that were not part of the constructed knowledge graph. Validation of the 443 hypotheses generated by the link prediction model resulted in 355 new food-chemical relationships, while results show that the model score correlates well (R2 = 0.70) with the probability of a novel finding. This work demonstrates how automated learning from literature at scale can accelerate discovery and support practical applications through reproducible, evidence-based capture of latent interactions of diverse entities, such as food and chemicals.

A novel link prediction model for interval-valued crude oil prices based on complex network and multi-source information

Accurate crude oil price forecasting is crucial for maintaining energy economy stability and enhancing investment and operational decision-making. Nowadays, the link prediction model for crude oil prices based on a complex network model has become an emerging and promising approach. However, existing link prediction methods are unable to extract complex information about interval-valued crude oil prices, and fail to consider the impact of multi-source information. Therefore, this paper proposes the link prediction model for interval-valued crude oil prices using complex network and multi-source information, which converts crude oil price interval series into complex networks. Through link prediction, it takes into account both the impact of past interval-crude oil prices on future interval time series and the interference of external real-time information on interval-valued prices. First, news headlines and geopolitical risk indices related to crude oil prices are captured, and search index data and news sentiment scores associated with interval data are analyzed. Subsequently, the interval sequence data and related influencing factors are decomposed and reconstructed respectively. The reconstructed interval center and radius sequence are then converted into a directed network, and node similarity in the network is calculated to generate similar nodes of the interval subsequence as feature values. Finally, multiple machine learning models are employed to acquire the optimal weighted combination prediction with interval prediction values. According to the experimental data, this approach outperforms other benchmark methods in terms of prediction accuracy.

Air route link prediction based on the PSO-CLP model

To optimize the structure of an air route network, accurate forecasting of future new routes is vital given the rapid growth in demand for air transportation. Based on the theory and method of link prediction, considering the joint influence of network endogenous factors and external attributes, we construct a system of network endogenous factors and external attribute indices and explore the prediction effect of each index. We further construct a prediction index system, explore the prediction effect of coupled indices, design a particle swarm algorithm to determine the weights of each index, and propose a coupled link prediction model based on particle swarm optimization (PSO-CLP). A comparison of the prediction accuracy of this model with the dual-indicator coupled link prediction model and the traditional link prediction model is also conducted to test the stability and reasonableness of the PSO-CLP model. In this study, we use the 2015–2020 Chinese air route network as an example. The instance test shows that the PSO-CLP models significantly outperform the traditional link prediction models and dual-indicator coupled link prediction models in terms of prediction accuracy, stability and computational simplicity, among which the PSO-CLP model, which considers both endogenous factors and external attributes, such as the RWR + Sor + Pop and RWR + RA + GDP indices, has the best forecasting effect. The PSO-CLP model is an effective tool for route prediction, providing a new perspective on route link prediction and air route network optimization.

Projection transformation-based link prediction for possible collaboration connections in bimodal supply chain networks

Aiming at the problems of information disconnection and business interruption in supply chain networks caused by uncertainties such as trade wars and COVID-19, a projection-transformation-based model for predicting possible collaborative relationship links in bimodal supply chain networks is designed from the perspective of network science. The model first transforms the bimodal network structure of supply chain into a corresponding unimodal projection network by projection transformation; then screens the candidate collaborative links in the network; finally assigns weights to the similarity links according to different influencing factors, and calculates the possibility of potential collaborative links in the bimodal supply chain network by this. The experimental results show that the model can achieve higher quality prediction in real bimodal supply chain networks compared with other baseline link prediction models, which provides theoretical and methodological support for better formulation of business contingency plans and risk defense measures.

MM-Transformer: A Transformer-Based Knowledge Graph Link Prediction Model That Fuses Multimodal Features

Multimodal knowledge graph completion necessitates the integration of information from multiple modalities (such as images and text) into the structural representation of entities to improve link prediction. However, most existing studies have overlooked the interaction between different modalities and the symmetry in the modal fusion process. To address this issue, this paper proposed a Transformer-based knowledge graph link prediction model (MM-Transformer) that fuses multimodal features. Different modal encoders are employed to extract structural, visual, and textual features, and symmetrical hybrid key-value calculations are performed on features from different modalities based on the Transformer architecture. The similarities of textual tags to structural tags and visual tags are calculated and aggregated, respectively, and multimodal entity representations are modeled and optimized to reduce the heterogeneity of the representations. The experimental results show that compared with the current multimodal SOTA method, MKGformer, MM-Transformer improves the Hits@1 and Hits@10 evaluation indicators by 1.17% and 1.39%, respectively, proving that the proposed method can effectively solve the problem of multimodal feature fusion in the knowledge graph link prediction task.

MTdyg: Multi-scale transformers with continuous time dynamic graph model for link prediction

Link prediction through continuous dynamic graph neural networks is a challenging endeavour. Previous studies have considered historic interaction sequences among pairs of nodes. However, this approach does not sufficiently model the links between them. Thus, the frequency of historical common neighbours was proposed to obtain more information about node pairs. Nonetheless, this measure fails to consider the timing of interactions because the relative importance of different interactions shifts over time.Multi-scale transformers with a continuous time dynamic graph model (MTdyg) are our response to this challenge to enhance link prediction. The novel MTdyg model employs a multi-scale patching strategy that dynamically adjusts based on the interaction frequency, feeding segmented interaction sequences into the transformer. This methodology significantly improves the model’s ability to assimilate historical data. Furthermore, we developed a temporal attention-based historical common neighbour encoding technique to effectively identify linkages between source and target nodes in interaction sequences. Extensive testing on eight public datasets confirmed that MTdyg delivers superior performance on most datasets, demonstrating its effectiveness in capturing the nuances of node interactions over time.

Analysis and prediction of the Horizon 2020 R&D&I collaboration network

This study presents a novel link prediction model for the analysis and prediction of collaboration networks in the context of H2020 projects. This model incorporates various databases and expands the list of variables used in the analysis. In this study, the accuracy and interpretability of both generic and nongeneric link prediction algorithms are examined through a comprehensive analysis. Additionally, collaborative groups that exhibit stronger linkages than initially anticipated are examined, and their prevalence across different regions is evaluated. In this study, it was found that the collaboration network within H2020 has significant concentration and fragmentation levels. Moreover, several key factors that strongly influence collaboration, including prior and concurrent project involvement, earned contribution, and the economic and technological characteristics of participating businesses, were identified. This study additionally demonstrated that cooperative communities exhibit geographical concentration in western Europe, whereas nongeneric models offer improved predictive accuracy for identifying relationships. This research contributes to the field of cooperation network analysis and offers guidance to decision-makers on strategies to promote future collaborations in the context of FP.

TLFSL: link prediction in multilayer social networks using trustworthy Lévy-flight semi-local random walk

Abstract As the landscape of online social networks continues to evolve, the task of expanding connections and uncovering novel relationships presents a growing complexity. Link prediction emerges as a crucial strategy, harnessing the current network dynamics to forecast future interactions among users. While traditional single-layer network link prediction models boast a storied legacy, recent attention has shifted towards tackling analogous challenges within the realm of multilayer networks. This paradigm shift underscores the critical role of extracting topological and multimodal features to effectively evaluate link weights, thereby enriching link prediction within weighted networks. Furthermore, the establishment of trustworthy pathways between users emerges as a pivotal tactic for translating unweighted similarities into meaningful weighted metrics. Leveraging the foundational principles of local random walk techniques, this paper introduces the trustworthy Lévy-flight semi-local (TLFSL) random walk framework for link prediction in multilayer social networks. By seamlessly integrating intralayer and interlayer information, TLFSL harnesses a dependable Lévy-flight random walk mechanism to anticipate new links within target layers of multilayer networks. Traditional local random walk techniques often overlook global relationships, as they confine path exploration to immediate neighbours. However, the absence of a direct edge between nodes does not necessarily imply a lack of relationship; nodes with semantic affinity may be spatially distant within the network. To overcome this limitation, we introduce the concept of semi-local random walk, which enables walker hopping with a wider global perspective. Meanwhile, TLFSL includes a distributed local community detection strategy to improve the performance of TLFSL in dealing with large-scale networks. Rigorous experimentation across diverse real-world multilayer networks consistently demonstrates TLFSL’s superior performance compared to equivalent methods.

A Deep Learning-Based Reliable Link Prediction Model for Achieving Traffic-Aware Routing in Mobile Ad-hoc Networks

One of the most promising wireless network architectures is the mobile ad-hoc network (MANET). Researchers have introduced enormous protocols for efficient routing, but it does not provide a reliable communication link for data transmission. Therefore, this research proposes a reliable link prediction-based traffic-aware deep learning routing protocol in MANET to maintain path stability and reliability to construct efficient routing. The reliable traffic-aware link prediction model used in this research is Fuzzy-based Deep Extreme [Formula: see text]-Learning (FDEQL) model. The fuzzy logic rule is used to compute the status of a wireless link to build stable and faster paths toward destinations. Traffic patterns can affect the efficiency of a node. So, to cope with the traffic pattern in MANET, the point-to-point (P2P) and end-to-end (E2E) traffic matrices are initially constructed. To evaluate whether the wireless link is reliable or not, the proposed approach utilizes fuzzy rules by considering essential parameters such as neighborhood overlap (NOVER), bipartivity index (BI), node mobility (NM), data rate (DR), received signal strength indicator (RSSI) and buffer occupancy (BO). The output is the [Formula: see text]-value for reliable link prediction. The performance of a proposed model is validated with other baseline methods based on various measures such as energy consumption, route failure, throughput, delay, packet delivery ratio (PDR), normalized routing load (NRL) and buffer occupancy by varying the mobility speed from 5 to 30[Formula: see text]m/sec, number of nodes and simulation time, respectively. At the mobility speed of 10[Formula: see text]m/sec, the proposed model has a delay of 0.08 sec, PDR of 99% and throughput of 1903.4[Formula: see text]kbps. The proposed model achieves a delay of 19.37[Formula: see text]msec, PDR of 96.22%, and throughput of 132.95[Formula: see text]kbps, respectively, for 30 nodes. If the simulation runs for 300 sec, the suggested model achieves a delay of 2.98 sec and a PDR of 0.946, respectively.

Reliable multiplex semi-local random walk based on influential nodes to improve link prediction in complex networks

In recent years, the exponential growth of online social networks as complex networks has presented challenges in expanding networks and forging new connections. Link prediction emerges as a crucial technique to anticipate future relationships among users, leveraging the current network state to address this challenge effectively. While link prediction models on monoplex networks have a well-established history, the exploration of similar tasks on multilayer networks has garnered considerable attention. Extracting topological and multimodal features for weighting links can improve link prediction in weighted complex networks. Meanwhile, establishing reliable and trustworthy paths between users is a useful way to create metrics that convert unweighted to weighted similarity. The local random walk is a widely used technique for predicting links in weighted monoplex networks. The aim of this paper is to develop a semi-local random walk over reliable paths to improve link prediction on a multilayer social network as a complex network, which is denoted as Reliable Multiplex semi-Local Random Walk (RMLRW). RMLRW leverages the semi-local random walk technique over reliable paths, integrating intra-layer and inter-layer information from multiplex features to conduct a trustworthy biased random walk for predicting new links within a target layer of multilayer networks. In order to make RMLRW scalable, we develop a semi-local random walk-based network embedding to represent the network in a lower-dimensional space while preserving its original characteristics. Extensive experimental studies on several real-world multilayer networks demonstrate the performance assurance of RMLRW compared to equivalent methods. Specifically, RMLRW improves the average f-measure of the link prediction by 3.2% and 2.5% compared to SEM-Path and MLRW, respectively.

ClusterLP: A novel Cluster-aware Link Prediction model in undirected and directed graphs

Link prediction models endeavor to understand the distribution of links within graphs and forecast the presence of potential links. With the advancements in deep learning, prevailing methods typically strive to acquire low-dimensional representations of nodes in networks, aiming to capture and retain the structure and inherent characteristics of networks. However, the majority of these methods primarily focus on preserving the microscopic structure, such as the first- and second-order proximities of nodes, while largely disregarding the mesoscopic cluster structure, which stands out as one of the network's most prominent features. Following the homophily principle, nodes within the same cluster exhibit greater similarity to each other compared to those from different clusters, suggesting that they should possess analogous vertex representations and higher probabilities of linkage. In this study, we develop a straightforward yet efficient Cluster-aware Link Prediction framework (ClusterLP), with the objective of directly leveraging cluster structures to predict links among nodes with maximum accuracy in both undirected and directed graphs. Specifically, we posit that establishing links between nodes with similar representation vectors and cluster tendencies is more feasible in undirected graphs, whereas nodes in directed graphs are inclined to point towards nodes with akin representation vectors and greater influence. We tailor the implementation of ClusterLP for undirected and directed graphs, respectively, and experimental findings using multiple real-world networks demonstrate the high competitiveness of our models in the realm of link prediction tasks. The code utilized in our implementation is accessible at https://github.com/ZINUX1998/ClusterLP.

A parameterised model for link prediction using node centrality and similarity measure based on graph embedding

Link prediction is a crucial aspect of graph machine learning, with applications as diverse as disease prediction, social network recommendations, and drug discovery. It involves the prediction of potential new links between nodes within a network. Despite its importance, current models for link prediction exhibit notable limitations. Graph Convolutional Networks have shown high efficiency in link prediction across various datasets. However, they face significant challenges when applied to short-path networks and ego networks, resulting in poor performance. This issue represents a critical area of concern that our work seeks to address. This paper introduces the Node Centrality and Similarity-based Parameterised Model (NCSM), a novel method for link prediction tasks. NCSM uniquely integrates node centrality and similarity measures as edge features in a customised Graph Neural Network (GNN) layer, effectively leveraging the topological information of large networks. This model represents the first parameterised GNN-based link prediction model that considers topological information. The proposed model was evaluated on five benchmark graph datasets, each comprising thousands of nodes and edges. Experimental results highlight NCSM's superiority over existing state-of-the-art models like Graph Convolutional Networks and Variational Graph Autoencoder, as it outperforms them across various metrics and datasets. This exceptional performance can be attributed to NCSM's innovative integration of node centrality, similarity measures, and its efficient use of topological information.

Repurposing non-pharmacological interventions for Alzheimer's disease through link prediction on biomedical literature

Non-pharmaceutical interventions (NPI) have great potential to improve cognitive function but limited investigation to discover NPI repurposing for Alzheimer's Disease (AD). This is the first study to develop an innovative framework to extract and represent NPI information from biomedical literature in a knowledge graph (KG), and train link prediction models to repurpose novel NPIs for AD prevention. We constructed a comprehensive KG, called ADInt, by extracting NPI information from biomedical literature. We used the previously-created SuppKG and NPI lexicon to identify NPI entities. Four KG embedding models (i.e., TransE, RotatE, DistMult and ComplEX) and two novel graph convolutional network models (i.e., R-GCN and CompGCN) were trained and compared to learn the representation of ADInt. Models were evaluated and compared on two test sets (time slice and clinical trial ground truth) and the best performing model was used to predict novel NPIs for AD. Discovery patterns were applied to generate mechanistic pathways for high scoring candidates. The ADInt has 162,212 nodes and 1,017,284 edges. R-GCN performed best in time slice (MR = 5.2054, Hits@10 = 0.8496) and clinical trial ground truth (MR = 3.4996, Hits@10 = 0.9192) test sets. After evaluation by domain experts, 10 novel dietary supplements and 10 complementary and integrative health were proposed from the score table calculated by R-GCN. Among proposed novel NPIs, we found plausible mechanistic pathways for photodynamic therapy and Choerospondias axillaris to prevent AD, and validated psychotherapy and manual therapy techniques using real-world data analysis. The proposed framework shows potential for discovering new NPIs for AD prevention and understanding their mechanistic pathways.