Link Prediction Problem Research Articles

Exploiting optimised communities in directed weighted graphs for link prediction

The most developing issue in analysing complex networks and graph mining is link prediction, which can be studied for both content and structural-based analysis in a social network. Link prediction deals with the prediction of missing links by determining whether a link can be created between two nodes in a future snapshot of a given directed weighted graph. Existing link prediction methods are only studied for unsigned graphs and work on principles of the common neighbourhood. However, the link prediction problem can also be studied for signed graphs where signed links can give an interesting insight into user associations. Obstruction of studies in this domain is caused by imbalance of class, i.e., positive links are frequent than negative ones, and forbearance of hidden communities. A signed network is a combination of dense and hidden communities. A hidden community structure is overlooked by majority of existing applications, taking dense community structure, i.e., one whole graph as input for developing a link prediction model. Hence, complete network information is required by majority of existing approaches, which seems unrealistic in modern social network analytics. In this article, we exploit hidden network communities to address link prediction problem in the signed network, focusing on negative links. A number of observation were made regarding negative links and a principle ensemble framework, i.e., ENeLp, is proposed, having two phases, i.e, network embedding and classifier prediction. Using a probabilistic embedding framework, network representation of hidden signed communities is learned, which were then passed to a learning classifier to predict negative links, keeping intact the ensemble framework. Despite the limited availability of signed network datasets, an extensive experimental study was performed to evaluate ENeLp pertinency, robustness, and scalability. The performance result shows that ENeLp can be a promising consideration for addressing link prediction tasks in signed networks and gives encouraging results.

MorbidGCN: prediction of multimorbidity with a graph convolutional network based on integration of population phenotypes and disease network.

Exploring multimorbidity relationships among diseases is of great importance for understanding their shared mechanisms, precise diagnosis and treatment. However, the landscape of multimorbidities is still far from complete due to the complex nature of multimorbidity. Although various types of biological data, such as biomolecules and clinical symptoms, have been used to identify multimorbidities, the population phenotype information (e.g. physical activity and diet) remains less explored for multimorbidity. Here, we present a graph convolutional network (GCN) model, named MorbidGCN, for multimorbidity prediction by integrating population phenotypes and disease network. Specifically, MorbidGCN treats the multimorbidity prediction as a missing link prediction problem in the disease network, where a novel feature selection method is embedded to select important phenotypes. Benchmarking results on two large-scale multimorbidity data sets, i.e. the UK Biobank (UKB) and Human Disease Network (HuDiNe) data sets, demonstrate that MorbidGCN outperforms other competitive methods. With MorbidGCN, 9742 and 14010 novel multimorbidities are identified in the UKB and HuDiNe data sets, respectively. Moreover, we notice that the selected phenotypes that are generally differentially distributed between multimorbidity patients and single-disease patients can help interpret multimorbidities and show potential for prognosis of multimorbidities.

A new multi-wave continuous action-set cellular learning automata for link prediction problem in weighted multi-layer social networks

A new multi-wave continuous action-set cellular learning automata for link prediction problem in weighted multi-layer social networks

Analysis and construction by convolution neural network of link prediction model on social network

Estimating similarity using multiple similarity measures or machine learning prediction models is a popular solution to the link prediction problem. The Relation Pattern Deep Learning Classification (RPDLC) technique is proposed in this study, and it is based on multiple neighbor-based similarity metrics and convolution neural networks. The RPDLC first calculates the characteristics for a pair of nodes using neighbor-based metrics and impact nodes. Second, the RPDLC creates a heat map using node characteristics to assess the similarity of the nodes’ connection patterns. Third, the RPDLC uses convolution neural network architecture to build a prediction model for missing relationship prediction. On three separate social network datasets, this method is compared to other state-of-the-art algorithms. On all three datasets, the suggested method achieves the greatest AUC, hovering around 99 percent. The use of convolution neural networks and features via relational patterns to create a prediction model are the paper’s primary contributions.

CFLP: A new cost based feature for link prediction in dynamic networks

Dynamic networks are social networks in which node-to-node links contain a temporal component, i.e., node-to-node interactions over a specific time interval. As a result, the dynamic network’s structure changes over time, and previously connected nodes may or may not have an edge connecting them at any one time. The link prediction issue in dynamic networks aims to identify future network linkages based on the relative behavior of previous network updates. We present a feature-based solution that considers both individual snapshots and the overall network throughout the full-time span to answer the link prediction problem. We present a novel feature called Cost-based feature for link prediction (CFLP) for estimating edge behavior throughout the entire network, which uses a reward and penalty structure to summarize node activity across the entire network. We use similarity indices, classified into four major categories: local similarity, global similarity, quasi-local similarity, and clustering coefficient-based similarity, to measure edge activity in individual snapshots. We have also selected fourteen different snapshot-based features to find the most excellent combination of minimum features for link prediction. We used regression and mutual information-based scoring for feature selection to correctly quantify the relative effect of features among themselves and the overall link prediction problem. In order to give the best feasible solution to the link prediction problem, these individual features and their combinations were examined with five machine learning models. We employed five performance matrices – AUC, AUPR, Average Precision, F1, and Balance Accuracy Score – to compare the performance of our method to those of state-of-the-art approaches, and found that our method outperformed all.

PWAF : Path Weight Aggregation Feature for link prediction in dynamic networks

The growing popularity of online social networks is obvious these days, and it provides researchers with an opportunity to find solutions for a variety of practical applications. The approach of recognizing network structure and finding missing and future links in a social network is known as link prediction. On time-varying or dynamic networks, the link prediction problem has two major challenges: accuracy and efficiency. The improvement of accuracy in link prediction problem in dynamic networks is the objective of this research. The goal of this paper is to propose the Path Weight Aggregation Feature (PWAF), which is a new feature based on ranking multi edge occurrences across the entire network. Different topological aspects of the networks (Local, Global, and Quasi−local) as well as Clustering Coefficient based features are taken into consideration for feature generation, in addition to the suggested Path Weight-Based Aggregation Feature (PWAF). One of the features used for better prediction is the Level−2 node clustering coefficient (CCLP2). Different machine learning models, such as Neural Network (NN), Logistic Regression (LR), XGBoost (XGB), Random Forest Classifier (RFC), and linear Discriminant Analysis (LDA), are evaluated and verified for link prediction. The experiments are carried out on seven different well-known dynamic networks data sets in terms of five performance evaluation metrics, including AUPR, F1-score, AVG PRECISION, BAL ACC SCORE, and AUC, and the results show that our proposed method and its variants outperform state-of-the-art methods. The results demonstrate that PWAF−RFC is the best performing variation out of all machine learning classifiers we have experimented with.

Multiple Order Local Information model for link prediction in complex networks

As a classical problem in the field of complex networks, link prediction has attracted much attention from researchers, which is of great significance to help us understand the evolution and dynamic development mechanisms of networks. Although various network type-specific algorithms have been proposed to tackle the link prediction problem, most of them suppose that the network structure is dominated by the Triadic Closure Principle. We still lack an adaptive and comprehensive understanding of network formation patterns for predicting potential links. In addition, it is valuable to investigate how network local information can be better utilized. To this end, we proposed a novel method named Link prediction using Multiple Order Local Information (MOLI) that exploits the local information from the neighbors of different distances, with parameter that can be a prior-driven based on prior knowledge, or data-driven by solving an optimization problem on observed networks. MOLI defined a local network diffusion process via random walks on the graph, resulting in better use of network information. We show that MOLI outperforms the other 12 widely used link prediction methods on 15 different types of simulated and real-world networks. We also conclude that there are different patterns of local information utilization for different networks, including social networks, communication networks, biological networks, etc. In particular, the classical common neighbor-based methods are not as adaptable to all social networks as it is perceived to be; instead, some of the social networks obey the Quadrilateral Closure Principle which preferentially connects paths of length three.

Predicting miRNA-disease associations via learning multimodal networks and fusing mixed neighborhood information.

In recent years, a large number of biological experiments have strongly shown that miRNAs play an important role in understanding disease pathogenesis. The discovery of miRNA-disease associations is beneficial for disease diagnosis and treatment. Since inferring these associations through biological experiments is time-consuming and expensive, researchers have sought to identify the associations utilizing computational approaches. Graph Convolutional Networks (GCNs), which exhibit excellent performance in link prediction problems, have been successfully used in miRNA-disease association prediction. However, GCNs only consider 1st-order neighborhood information at one layer but fail to capture information from high-order neighbors to learn miRNA and disease representations through information propagation. Therefore, how to aggregate information from high-order neighborhood effectively in an explicit way is still challenging. To address such a challenge, we propose a novel method called mixed neighborhood information for miRNA-disease association (MINIMDA), which could fuse mixed high-order neighborhood information of miRNAs and diseases in multimodal networks. First, MINIMDA constructs the integrated miRNA similarity network and integrated disease similarity network respectively with their multisource information. Then, the embedding representations of miRNAs and diseases are obtained by fusing mixed high-order neighborhood information from multimodal network which are the integrated miRNA similarity network, integrated disease similarity network and the miRNA-disease association networks. Finally, we concentrate the multimodal embedding representations of miRNAs and diseases and feed them into the multilayer perceptron (MLP) to predict their underlying associations. Extensive experimental results show that MINIMDA is superior to other state-of-the-art methods overall. Moreover, the outstanding performance on case studies for esophageal cancer, colon tumor and lung cancer further demonstrates the effectiveness of MINIMDA. https://github.com/chengxu123/MINIMDA and http://120.79.173.96/.

GraphTGI: an attention-based graph embedding model for predicting TF-target gene interactions

Interaction between transcription factor (TF) and its target genes establishes the knowledge foundation for biological researches in transcriptional regulation, the number of which is, however, still limited by biological techniques. Existing computational methods relevant to the prediction of TF-target interactions are mostly proposed for predicting binding sites, rather than directly predicting the interactions. To this end, we propose here a graph attention-based autoencoder model to predict TF-target gene interactions using the information of the known TF-target gene interaction network combined with two sequential and chemical gene characters, considering that the unobserved interactions between transcription factors and target genes can be predicted by learning the pattern of the known ones. To the best of our knowledge, the proposed model is the first attempt to solve this problem by learning patterns from the known TF-target gene interaction network. In this paper, we formulate the prediction task of TF-target gene interactions as a link prediction problem on a complex knowledge graph and propose a deep learning model called GraphTGI, which is composed of a graph attention-based encoder and a bilinear decoder. We evaluated the prediction performance of the proposed method on a real dataset, and the experimental results show that the proposed model yields outstanding performance with an average AUC value of 0.8864 +/- 0.0057 in the 5-fold cross-validation. It is anticipated that the GraphTGI model can effectively and efficiently predict TF-target gene interactions on a large scale. Python code and the datasets used in our studies are made available at https://github.com/YanghanWu/GraphTGI.

A Novel Multilayer Model for Link Prediction in Online Social Networks Based on Reliable Paths

Link Prediction (LP) is one of the critical problems in Online Social Networks (OSNs) analysis. LP is a technique for predicting forthcoming or missing links based on current information in the OSN. Typically, modelling an OSN platform is done in a single-layer scheme. However, this is a limitation which might lead to incorrect descriptions of some real-world details. To overcome this limitation, this paper presents a multilayer model of OSN for the LP problem by analysing Twitter and Foursquare networks. LP in multilayer networks involves performing LP on a target layer benefitting from the structural information of the other layers. Here, a novel criterion is proposed, which calculates the similarity between users by forming intralayer and interlayer links in a two-layer network (i.e. Twitter and Foursquare). Particularly, LP in the Foursquare layer is done by considering the two-layer structural information. In this paper, according to the available information from the Twitter and Foursquare OSNs, a weighted graph is created and then various topological features are extracted from it. Based on the extracted features, a database with two classes of link existence and no link has been created, and therefore the problem of LP has become a two-class classification problem that can be solved by supervised learning methods. To prove the better performance of the proposed method, Katz and FriendLink indices as well as SEM-Path algorithm have been used for comparison. Evaluations results show that the proposed method can predict new links with better precision.

Complex-InversE: Improving bilinear knowledge graph embeddings by mapping into complex space

Knowledge graph link prediction uses known fact links to infer the missing link information in the knowledge graph, which is of great significance to the completion of the knowledge graph. Generating low-dimensional embeddings of entities and relations which are used to make inferences is a popular way for such link prediction problems. This paper proposes a knowledge graph link prediction method called Complex-InversE in the complex space, which maps entities and relations into the complex space. The composition of complex embeddings can handle a large variety of binary relations, among them symmetric and antisymmetric relations. The Complex-InversE effectively captures the antisymmetric relations and introduces Dropout and Early-Stopping technologies into deal with the problem of small numbers of relationships and entities, thus effectively alleviates the model’s overfitting. The results of comparison experiment on the public knowledge graph datasets show that the Complex-InversE achieves good results on multiple benchmark evaluation indicators and outperforms previous methods. Complex-InversE’s code is available on GitHub at https://github.com/ZeyuMiao97/Complex-InversE.

Graph regularized nonnegative matrix factorization for link prediction in directed temporal networks using PageRank centrality

Temporal link prediction aims to predict new links at time T + 1 based on a given temporal network from time 1 to T. Some existing temporal link prediction methods only consider the naturally links direction formation but ignore the local and global information of temporal networks. To address this issue, in this paper, we propose a novel graph regularized nonnegative matrix factorization algorithm for the problem of temporal link prediction by using PageRank centrality and asymmetric link clustering coefficient, referred to as the GNMFCA. Specifically, we use graph regularization to capture the local information of each slice in temporal networks and PageRank centrality to compute the importance of nodes in each slice, which captures the global information of each slice in temporal networks. By jointly optimizing them in a nonnegative matrix factorization model, the proposed model can preserve both the local and global information simultaneously. Besides, we propose effective multiplicative updating rules to solve the proposed model and provide the convergence analysis of the iterative algorithm. Finally, we perform numerical experiments on eight artificial and four real-world temporal networks to demonstrate that GNMFCA outperforms some existing temporal link prediction algorithms.

FLP-ID: Fuzzy-based link prediction in multiplex social networks using information diffusion perspective

The growing popularity of online social networks is evident nowadays and allows researchers to find solutions for various practical applications. Link prediction is the technique of understanding network structure and identifying the missing links in the social network. The two significant challenges of the link prediction problem are accuracy and efficiency on growing and multiplex networks. Well-known methods for link prediction are the similarity-based methods, which use local, global, and topological features of the network to predict missing links. These approaches ignore critical factors such as different channels of interaction, information diffusion, group norms to form new connections. Therefore, a fuzzy-based link prediction algorithm (FLP-ID) in multiple social networks is proposed using information diffusion. First, FLP-ID generates a multiplex network by combining different types of relationships among users and identifying the community structure. Thereafter, the algorithm computes node and relative relevance for distinct fuzzy criteria under group norms. Finally, the likelihood score of each non-existing link is computed to predict missing links. The experimental results show that the proposed fuzzy algorithm accuracy is better than crisp algorithms over the multiplex network. The prediction rate of FLP-ID with F1-score, AUC, and balanced accuracy is excellent, which are improved compared to related methods up to 30%, 35%, and 30%, respectively, on high density and clustering coefficient datasets under multiplex settings.

DTI-HETA: prediction of drug-target interactions based on GCN and GAT on heterogeneous graph.

Drug-target interaction (DTI) prediction plays an important role in drug repositioning, drug discovery and drug design. However, due to the large size of the chemical and genomic spaces and the complex interactions between drugs and targets, experimental identification of DTIs is costly and time-consuming. In recent years, the emerging graph neural network (GNN) has been applied to DTI prediction because DTIs can be represented effectively using graphs. However, some of these methods are only based on homogeneous graphs, and some consist of two decoupled steps that cannot be trained jointly. To further explore GNN-based DTI prediction by integrating heterogeneous graph information, this study regards DTI prediction as a link prediction problem and proposes an end-to-end model based on HETerogeneous graph with Attention mechanism (DTI-HETA). In this model, a heterogeneous graph is first constructed based on the drug-drug and target-target similarity matrices and the DTI matrix. Then, the graph convolutional neural network is utilized to obtain the embedded representation of the drugs and targets. To highlight the contribution of different neighborhood nodes to the central node in aggregating the graph convolution information, a graph attention mechanism is introduced into the node embedding process. Afterward, an inner product decoder is applied to predict DTIs. To evaluate the performance of DTI-HETA, experiments are conducted on two datasets. The experimental results show that our model is superior to the state-of-the-art methods. Also, the identification of novel DTIs indicates that DTI-HETA can serve as a powerful tool for integrating heterogeneous graph information to predict DTIs.

Link Prediction in Time Varying Social Networks

Predicting new links in complex networks can have a large societal impact. In fact, many complex systems can be modeled through networks, and the meaning of the links depend on the system itself. For instance, in social networks, where the nodes are users, links represent relationships (such as acquaintance, friendship, etc.), whereas in information spreading networks, nodes are users and content and links represent interactions, diffusion, etc. However, while many approaches involve machine learning-based algorithms, just the most recent ones account for the topology of the network, e.g., geometric deep learning techniques to learn on graphs, and most of them do not account for the temporal dynamics in the network but train on snapshots of the system at a given time. In this paper, we aim to explore Temporal Graph Networks (TGN), a Graph Representation Learning-based approach that natively supports dynamic graphs and assigns to each event (link) a timestamp. In particular, we investigate how the TGN behaves when trained under different temporal granularity or with various event aggregation techniques when learning the inductive and transductive link prediction problem on real social networks such as Twitter, Wikipedia, Yelp, and Reddit. We find that initial setup affects the temporal granularity of the data, but the impact depends on the specific social network. For instance, we note that the train batch size has a strong impact on Twitter, Wikipedia, and Yelp, while it does not matter on Reddit.

Joint Learning of User Representation With Diffusion Sequence and Network Structure

Information sharing behavior and social link building behavior have shown strong correlation on social media. The aim of this paper is to explore this correlation for simultaneously modeling and predicting sharing behavior in information diffusion sequences and linking behavior in social network, which correspond to information diffusion prediction and social link prediction problems. To achieve this goal, we propose a joint user representation learning model to characterize the two correlated behaviors in a shared latent space. The proposed model learns user representations via two maximum likelihood estimation objectives defined on observed information diffusion sequences and social network structure respectively and incorporates them in a unified framework. A multi-task learning algorithm is designed for efficient model optimization. Based on the learned representations, the model can be directly applied to predicting diffusion processes and inferring unobserved social links at the same time. We evaluate the proposed model on two real social media datasets with extensive experiments. The model consistently achieves significant improvements over the state-of-the-art approaches on diffusion prediction and link prediction tasks. The better robustness of our model in further ablation studies demonstrates that capturing the behavior correlation in the shared representation space is beneficial.

Link prediction for long-circle-like networks.

Link prediction is the problem of predicting the uncertain relationship between a pair of nodes from observed structural information of a network. Link prediction algorithms are useful in gaining insight into different network structures from partial observation of exemplars. Existing local and quasilocal link prediction algorithms with low computational complexity focus on regular complex networks with sufficiently many closed triangular motifs or on tree-like networks with the vast majority of open triangular motifs. However, the three-node motif cannot describe the local structural features of all networks, and we find the main structure of many networks is long line or closed circle that cannot be predicted well via traditional link prediction algorithms. Meanwhile, some global link prediction algorithms are effective but accompanied by high computational complexity. In this paper, we proposed a local method that is based on the natural characteristic of a long line-in contrast to the preferential attachment principle. Next, we test our algorithms for two kinds of symbolic long-circle-like networks: a metropolitan water distribution network and a sexual contact network. We find that our method is effective and performs much better than many traditional local and global algorithms. We adopt the community detection method to improve the accuracy of our algorithm, which shows that the long-circle-like networks also have clear community structure. We further suggest that the structural features are key for the link prediction problem. Finally, we propose a long-line network model to prove that our core idea is of universal significance.

Predicting future links with new nodes in temporal academic networks

Most real-world systems evolve over time in which entities and the interactions between entities are added and removed—new entities or relationships appear and old entities or relationships vanish. While most network evolutionary models can provide an iterative process for constructing global properties, they cannot capture the evolutionary mechanisms of real systems. Link prediction is hence proposed to predict future links which also can help us understand the evolution law of real systems. The aim of link prediction is to uncover missing links from known parts of the network or quantify the likelihood of the emergence of future links from current structures of the network. However, almost all existing studies ignored that old nodes tend to disappear and new nodes appear over time in real networks, especially in social networks. It is more challenging for link prediction since the new nodes do not have pre-existing structure information. To solve the temporal link prediction problems with new nodes, here we take into account nodal attribute similarity and the shortest path length, namely, ASSPL, to predict future links with new nodes. The results tested on scholar social network and academic funding networks show that it is highly effective and applicable for ASSPL in funding networks with time-evolving. Meanwhile, we make full use of an efficient parameter to exploit how network structure or nodal attribute has an impact on the performance of temporal link prediction. Finally, we find that nodal attributes and network structure complement each other well for predicting future links with new nodes in funding networks.

Link prediction based on higher-order structure extraction and autoencoder learning in directed networks

Link prediction, which aims to estimate the likelihood of links based on the observed links and/or node attributes, is a common task in complex network analysis and graph data mining. Modeling and exploring high-order structure, also called network motif, are essential for understanding the fundamental structure that control and mediate the behavior of various complex systems. Most existing link prediction methods do not fully consider the high-order structure in the network. In addition, they mainly focus on undirected networks and therefore ignore the directionality of links. To this end, we propose a novel Autoencoder model based on high-order structure to solve the problems of existing link prediction methods. Specifically, we first provide an efficient motif adjacency matrix learning algorithm, which can be used to extract high-order structure in directed networks, and construct multiple motif adjacency matrices. Then, we extend the Graph Autoencoder (GAE) and Variational Autoencoder (VAE) frameworks to solve the link prediction problem in directed networks, which consists of three core modules. (i) First (Node Embedding): a novel encoder scheme is presented to learn the node embeddings from each motif adjacency matrix; (ii) Second (Information Fusion): a robust attention scheme is further introduced to aggregate information of node embeddings learned from multiple motif adjacency matrices to form the final semantically rich node embedding matrix; (iii) Third (Link Prediction): an efficient decoder scheme is finally proposed to reconstruct the target directed network in a multiscale way, by jointly utilizing node embeddings and PageRank centrality, the directions of links can be effectively inferred based on the classic gravitational heuristic formula. Extensive experiments were applied on various types of directed networks to validate the performance of our proposed approach through comparisons with the state-of-the-art link prediction technologies.

A novel similarity measure of link prediction in multi‐layer social networks based on reliable paths

AbstractOnline social networks are an integral element of modern societies and significantly influence the formation and consolidation of social relationships. In fact, these networks are multi‐layered so that there may be multiple links between a user' on different social networks. In this article, the link prediction problem for the same user in a two‐layer social network is examined, where we consider Twitter and Foursquare networks. Here, information related to the two‐layer communication is used to predict links in the Foursquare network. Link prediction aims to discover spurious links or predict the emergence of future links from the current network structure. There are many algorithms for link prediction in unweighted networks, however only a few have been developed for weighted networks. Based on the extraction of topological features from the network structure and the use of reliable paths between users, we developed a novel similarity measure for link prediction. Reliable paths have been proposed to develop unweight local similarity measures to weighted measures. Using these measures, both the existence of links and their weight can be predicted. Empirical analysis shows that the proposed similarity measure achieves superior performance to existing approaches and can more accurately predict future relationships. In addition, the proposed method has better results compared to single‐layer networks. Experiments show that the proposed similarity measure has an advantage precision of 1.8% over the Katz and FriendLink measures.