Open research questions in Complex Network Analysis Techniques

What the literature leaves open

• this graph neural networks, This paper proposes an innovative framework community detection with integrating dynamic effectively addressing the trajectory prediction, challenges of data sparsity and missing social context encountered by traditional methods when modelling specific individuals' behavioural trajectories. By incorporating community evolution analysis from dynamic networks into trajectory modelling, and combining bidirectional neural time-point processes with approach simultaneously captures both the spatio-temporal characteristics of individual behaviour and the constraints imposed by their social relationships.

  Precise Modelling Method for Specific Individual Behavioural Trajectories Based on Community Detection · 2026 · DOI

• prediction accuracy three core the for Looking ahead, several directions for future expansion remain.

  Precise Modelling Method for Specific Individual Behavioural Trajectories Based on Community Detection · 2026 · DOI

• The Proposed method consistently performed better across various benchmark net- works, including structurally dense networks such as Caltech and Polblog. By combining Graph Diffusion Convolution, local and global representation learning were enhanced, leading to improved and stable link prediction results. The ablation experiment revealed that AUC was improved by up to 19%, Precision by up to 25.3%, and F1-score by up to 29.1% when compared to GraphSAGE alone. These findings show that GDC significantly improves graph connectivity and representation learning, resulting in high accuracy, robustness, and generalization in link prediction across different network structures. Future work will focus on extending the proposed framework to fully dynamic graph learning settings via incremental diffusion updates, temporal node embeddings, and time-aware graph neural architectures that model continuously evolving network interactions.

  Enhancing link prediction in complex networks using GraphSAGE with graph diffusion convolution · 2026 · DOI

• A Receiver Operating Characteristic (ROC) curve is a visual representation illustrating the efficacy of a method in discerning true positive samples and differentiating them from negative samples. This graphical plot depicts the true-positive rate plotted against the falsepositive rate across various thresholds, thereby delineating the performance characteristics of the method across a range of discrimination thresholds. Figure 7 depicts the ROC curves for each network, which assesses the effectiveness of the proposed approach in comparison to other methods being evaluated. The SLC-GCN has surpassed all other approaches, including local, semi-local, global, and state-of-the-art methods, in nearly all datasets and has achieved the highest area under the curve. These curves demonstrate that utilizing mutual influence for calculating edge weights can significantly enhance link prediction ability. Figure 8 depicts the impact of varying training sizes on the performance of the proposed approach compared to existing algorithms. As illustrated, it is evident that, overall, the accuracy of prediction improves as the training size increases. SLC-GCN consistently outperforms local, semi-local, global, and state-of-the-art methods in terms of AUC across various training set sizes in almost all datasets. This is of great significance as it demonstrates that even with limited access to a portion of observed edges, the SLC-GCN is capable of accurately predicting the non-observed edges, surpassing the accuracy of current cutting-edge approaches. Specifically, in the majority of networks, when the training size is extremely limited, the proposed approach exhibits a notable superiority compared to the local metrics. In the networks of FOOTBALL, CELEGANS, and NETSCIENCE, SLC-GCN demonstrates an accuracy that is 8.4%, 11.3%, and 10.6% higher than that of local metrics. When comparing with other semi-local based metrics such as FL and LRW, the AUC improvement for SLC-GCN is roughly 6.3%, 4.9%, and 6.8% for the FOOTBALL, CELEGANS, and NETSCIENCE networks, respectively. This highlights the importance of various topological and multimodal features in evaluating similarity through random walks. Also, SLC-GCN is also superior compared to global metrics and state-of-the-art methods. On average, SLC-GCN outperforms KI and SEM-Paths by 4.7% and 1.3%, respectively.

  Community detection in complex networks via semi-supervised learning: application of link prediction · 2026 · DOI

• • Although our discussions involve concepts related to bullying and victimization, the data are limited to students’ reports of positive and negative relationships. As such, the study cannot directly identify bullying behaviors, but rather captures structural patterns of negative social relationships. • As we mentioned before, since our analyses are conditioned on subnetworks constructed from the most negatively con- nected nodes, some boundary effects may arise. Additionally, the subnetwork construction conditions on the selection of negative hubs, which may introduce structural biases. By focusing only on nodes with high negative degree, the result- ing subnetworks reflect a subset of the social system, and their properties may differ from those of the full network. • The data were collected from schools in a single region of México, which may limit the generalizability of the findings to other cultural or institutional contexts. • The cross-sectional design does not allow us to examine temporal dynamics or make causal claims regarding the emer- gence of negative or positive ties. Longitudinal data would be necessary to evaluate how attack and defense relations evolve over time.

  Attack and defense networks in a student social system · 2026 · DOI

• The prediction of the correlation coefficient of the knowledge label network by the ARIMA model shows that the connection between the labels is lacking diversity and the opinion strengthening phenomenon tends to strengthen, which is more likely to form the “echo chamber effect”, resulting in mutual isolation and even opposition between different circles.

  Analysis of the characteristics and evolution of knowledge label networks in the Q&A community: taking the Zhihu platform as an example · 2023 · DOI

• Purpose The state-of-the-art methods designed for overlapping community detection are limited by their high execution time as in CPM or the need to provide some parameters like the number of communities in Bigclam and Nise_sph, which is a nontrivial information.

  Overlapping community detection based on the union of all maximum spanning trees · 2020 · DOI

• Although it is well known that some exponential family random graph model (ERGM) families exhibit phase transitions (in which small parameter changes lead to qualitative changes in graph structure), the behavior of other models is still poorly understood.

  Phase transitions in the edge/concurrent vertex model · 2020 · DOI

• In particular, adopting the (asymptotic) scaling of the variance of the maximum likelihood parameter estimates as a notion of effective sample size ($n_{\mathrm{eff}}$), we show that when modeling the overall propensity to have ties and the propensity to reciprocate ties, whether the networks are sparse or not under the model (i.

  On the Question of Effective Sample Size in Network Modeling: An Asymptotic Inquiry · 2015 · DOI

• However, despite various theoretical advancements, the CCO perspective’s range of methodologies is still limited to analyzing local communication episodes, rather than studying organizations as broader networks of communication episodes.

  Organizations as Networks of Communication Episodes: Turning the Network Perspective Inside Out · 2012 · DOI

• The areal random growth models introduced successfully simulate many of the characteristics of natural stream systems, but it remains uncertain to what extent inferences may be drawn about the processes responsible for natural stream systems on the basis of this intriguing correspondence.

  Simulation of Stream Networks by Headword Growth and Branching* · 1971 · DOI

• This being said, we expect interesting developments in future years on the open problem of the optimal choice of path, where applications will be a driving force in this regard.

  Global Analysis of Regulatory Network Dynamics: Equilibria and Saddle-Node Bifurcations · 2026 · DOI

• 's inter-order overlap parameter is insufficient alone to express the model within our framework despite performing well against simulations, with the original derivation implicitly invoking additional structural assumptions.

  A principled closure framework for higher-order SIS epidemic models on networks · 2026

• This effect stems from preferential attachment term predicting more than one edge between some pairs of nodes and is only limited to the few largest in-degree nodes.

  The Price–Pareto growth model of networks with community structure · 2026 · DOI

• A key limitation of the present validation is that the case studies are retrospective and no controlled non-UNMF baseline is provided.

  Networked data science: a unified network modeling framework · 2026 · DOI

• However, despite existing work on node-removal vulnerabilities, little is known about multipartite robustness when attackers lack precise network information and thus exhibit imperfect target discrimination.

  Robustness of multipartite networks under attacks with imperfect target discrimination · 2026 · DOI

• Link-centred and binary percolation measures identify important facilities or connectivity failures, but they provide limited information on which spatial areas cause the largest loss of network-wide propagation capability.

  Quantum percolation based dynamic propagation connectivity for critical-area identification in transport networks · 2026

• Despite these open questions, this article’s results show that the systematic and standardized analysis of external data sources is feasible and methodologically sound, and represents an important step towards more data-driven and proactive risk management in modern companies.

  A model for high-frequency detection of current risks based on news analysis and decentralized social networks · 2026 · DOI

• Although social capital is widely recognized as a key determinant of resilience, its dynamic restructuring after disruption remains poorly quantified.

  Disaster-induced behavioral change restructures social networks toward bonding ties · 2026

• Abstract The emergence of hypergraphs has solved the problem that the interactions between nodes are insufficient to describe the complex relationships among multiple individuals.

  Message passing method for social contagion in hypergraphs · 2026 · DOI

• Open questions Six questions remain open: 1) Synthetic Layer 2. As a single-body observation, the graph-shear mechanism warrants further investigation across bodies with comparable rotation- orbit commensurability.

  From Classification to Mechanism in Temporal Contact Networks · 2026 · DOI

• Although higher budgets expand the number of iterations proportionally, the practical availability of rewiring opportunities remains limited by the overall budget capacity.

  Novel rewiring mechanism for restoration of the fragmented social networks after attacks · 2026 · DOI

• Despite the recent interest in higher-order networks, little is known about the mechanisms that govern the formation and evolution of groups, and how people move between groups.

  The temporal dynamics of group interactions in higher-order social networks · 2024 · DOI

• Abstract Despite the growing interest in interorganizational border management, relatively little is known about antecedents that drive such coordination efforts emerging in and around border regions.

  Antecedents of Border Management Network in El Paso, Texas: An Exponential Random Graph Model · 2018 · DOI

• By considering the population as a bipartite graph of a two-mode network (those from the sampling frame and those who are not on the frame), the number of respondents who are directly linked to the sampling frame members can be estimated using Chao’s and Zelterman’s estimators for sparse data.

  B-Graph Sampling to Estimate the Size of a Hidden Population · 2015 · DOI

Most-cited papers in Complex Network Analysis Techniques

• Interorganizational Networks at the Network Level: A Review of the Empirical Literature on Whole Networks · Journal of Management · 2007 · 1,020 citations
• Alone in the crowd: The structure and spread of loneliness in a large social network. · Journal of Personality and Social Psychology · 2009 · 620 citations
• Evolutionary dynamics of higher-order interactions in social networks · Nature Human Behaviour · 2021 · 545 citations
• Dynamic Network Visualization · American Journal of Sociology · 2005 · 318 citations
• Centrality measures in networks · Social Choice and Welfare · 2023 · 154 citations
• Epidemic spreading on higher-order networks · Physics Reports · 2024 · 152 citations
• Social Network Analysis: An Example of Fusion Between Quantitative and Qualitative Methods · Journal of Mixed Methods Research · 2018 · 116 citations
• Organizations as Networks of Communication Episodes: Turning the Network Perspective Inside Out · Organization Studies · 2012 · 99 citations
• A network perspective on suicidal behavior: Understanding suicidality as a complex system · Suicide and Life-Threatening Behavior · 2021 · 96 citations
• Network Dynamics and Organizations: A Review and Research Agenda · Journal of Management · 2022 · 92 citations

Most recent work

• Detecting communities when order and direction matter in social network analysis · Canadian Journal of Statistics · 2026
• Using covariance of node states to design early warning signals for network dynamics · Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences · 2026
• Ecological networks balance connectivity with flexibility to perturbations · bioRxiv · 2026
• Contagion of Negative Emotion in Structurally Heterogeneous Social Networks: A Cascade-Based Analysis · Chinese Physics B · 2026
• Binary Relations in Friendship Networks: A Mathematical Framework for Reflexivity, Symmetry, Transitivity and Community Formation · Asian Research Journal of Mathematics · 2026
• Community correlations and testing independence between binary graphs · Applied Network Science · 2026
• Message passing method for social contagion in hypergraphs · Chinese Physics B · 2026
• A statistical test for network similarity · Journal of Physics: Complexity · 2026
• Matrix Methods for Connectivity and Reachability in Transportation Networks · Asian Research Journal of Mathematics · 2026
• Whale optimization algorithm based on Markov chain is used for overlapping community discovery · Scientific Reports · 2026

Related topics in Physics and Astronomy