longitudinal network analysis

Addressing core challenges in relational event network modeling—including difficult model evaluation, weak theoretical foundations, lack of principled intervention design, and low predictive interpretability—this paper introduces the first statistical simulation framework specifically designed for relational event networks. It integrates binary dependence structures with actor-oriented mechanisms to enable fine-grained, temporally resolved Monte Carlo generation of relational events. Methodologically, it is the first work to systematically unify relational event simulation with social network analysis across four key application scenarios: model diagnostics, theory testing, intervention forecasting, and structural attribution. We release an open-source R package, *remulate*, ensuring reproducibility and interpretability. Through three empirical case studies—criminal gang evolution, validation of optimal distinctiveness theory, and network intervention analysis—we demonstrate the framework’s effectiveness in enhancing model discriminability, advancing theoretical development, and quantifying causal intervention effects.

Static snapshot-based approaches fail to capture temporal information flow and the evolution of propagation paths in time-evolving graphs. To address this, we propose the first temporal path tracing system designed for highly dynamic vertices. Our method abandons the conventional static graph snapshot paradigm and instead supports configurable, time-constrained path traversal algorithms, integrated with parallel graph processing and web-based interactive visualization. At the modeling level, we adopt an exact temporal graph representation that preserves fine-grained temporal semantics. Computationally, we enable efficient, low-latency path queries over large-scale evolving graphs. Analytically, our system significantly improves both the accuracy and real-time responsiveness of critical propagation path identification. Empirical evaluation demonstrates that our system outperforms state-of-the-art methods in both query efficiency and interpretability, establishing a new benchmark for temporal path analysis in dynamic networks.

This study identifies a critical reproducibility crisis in keyword co-occurrence network analysis, stemming from the high sensitivity of analytical outcomes to preprocessing decisions—including keyword extraction methods, co-occurrence definitions (e.g., document- vs. sentence-level), and filtering thresholds—exacerbated by widespread lack of transparency in network construction protocols. Through systematic multi-scenario experiments on bibliometric and social media datasets, we quantitatively assess how alternative preprocessing pathways perturb key network topology metrics: degree centrality, clustering coefficient, and modularity exhibit fluctuations exceeding 200% under varying configurations. To address this, we introduce the “data provenance transparency” framework, advocating standardized, granular documentation of all network construction decisions. Our findings deliver both a theoretical caution against methodological opacity in complex network science and concrete, actionable guidelines for enhancing analytical credibility and interpretability.

Traditional process analysis relies on static, aggregated data, failing to capture the causal dynamics of individual behavior evolution over time. This paper proposes a novel method integrating temporal behavioral modeling with Granger causality inference. Specifically, it is the first to combine fine-grained temporal patterns of actor interactions with Group Lasso-based lag selection to automatically identify key lagged behavioral structures exerting persistent causal effects on process performance metrics—such as throughput time—directly from event logs. The approach breaks away from prevailing static assumptions about behavior–outcome relationships. Empirical evaluation on real-world industrial data reveals statistically significant and quantifiable cross-temporal causal impacts of individual behaviors—including handover delays and task interruptions—on downstream process outcomes. By uncovering interpretable, intervention-ready behavioral-level causal mechanisms, the method provides a rigorous foundation for data-driven process optimization.

Novice learners in network science lack systematic, interdisciplinary guidance. Method: We propose the “Network Scientist Training Map” framework, integrating diverse methodological tools—including graph theory, random graph models, spectral graph theory, network embedding, statistical inference, dynamical modeling, and interpretable machine learning—restructuring core paradigms in an accessible, unified manner that emphasizes methodological synthesis over technical accumulation. Contribution/Results: First, we formally establish pure network science as an autonomous academic discipline. Second, we construct a structured, full-stack knowledge graph covering foundational concepts and techniques. Third, by transcending the limitations of single-discipline textbooks, the framework enables learners with no prior background to rapidly develop a coherent, systems-level conceptual framework—thereby advancing network science from an application-oriented auxiliary field toward ontological independence.

This study addresses the dynamic identification and robustness assessment of key actors in criminal networks. To tackle challenges posed by missing data and underutilized temporal information in judicial records, we propose a time-aware dynamic Katz centrality measure, coupled with a Bernoulli-based probabilistic network completion model. This integration enables modeling of individual centrality evolution over time and distinguishes long-term core members from transiently influential participants. The method ensures stability under data uncertainty, and empirical evaluation confirms that individuals exhibiting persistently high centrality serve as critical organizational hubs. To our knowledge, this is the first work to jointly leverage dynamic centrality and probabilistic network imputation for analyzing incomplete criminal networks—establishing a novel, interpretable, and robust paradigm for key-node identification.

为了解决动态属性图中链接预测和社区发现的问题，提出了一种新的信息工作流程inc-LPCDAG，结合结构和属性信息分析社交系统。

Modeling time-varying spillover effects, autolag persistence, and covariate impacts in network-structured time series remains challenging due to the entanglement of topology and temporal dynamics. Method: We propose the Network Time-Varying Parameter Vector Autoregression (NTVP-VAR) model, which represents lag matrices as time-varying linear combinations of graph operators—thereby decoupling network structure (interaction topology) from temporal evolution (interaction strength). It employs low-dimensional latent states to govern dynamic node interactions, autolags, and covariate responses, integrating graph differencing, stochastic coefficient evolution, and tensor low-rank constraints. Contribution/Results: Theoretically, we establish conditions for second-moment existence, network stability, and local stationarity under nonstationary latent states. Methodologically, NTVP-VAR unifies Gaussian/Poisson network VARs, graph-differenced ARIMA, and dynamic edge models. Empirically, it achieves superior predictive accuracy and interpretability on high-dimensional sparse networks.

This paper addresses graph representation learning for both static and single-event dynamic networks. Methodologically, it introduces a unified structural-aware embedding framework grounded in latent distance modeling, jointly optimizing homophily, transitivity, and balance within an end-to-end paradigm—thereby eliminating heuristic design and multi-stage pipelines. Notably, it is the first to extend latent distance modeling to single-event dynamic settings, enabling extreme node identification and quantitative assessment of influence dynamics. The key contributions are: (1) a hierarchical, interpretable structural-aware representation; (2) seamless unification of embedding learning across static and dynamic networks; and (3) state-of-the-art performance on community detection, anomaly detection, and temporal influence evaluation—significantly outperforming multi-stage baselines.

This study addresses the challenge of applying conventional statistical methods to collections of heterogeneous networks that vary in size and type and lack node correspondence. To overcome this, the authors propose a functional Topological Data Analysis (funTDA) framework that uniquely integrates functional data analysis with persistent homology to extract topological features from networks. This approach enables standard statistical operations—including mean and variance estimation, principal component analysis, and hypothesis testing—despite the non-Euclidean nature of network structures, thereby establishing a unified inferential framework. Empirical evaluations demonstrate that funTDA effectively discriminates networks with distinct connectivity patterns and successfully uncovers significant topological differences in real-world applications, such as literary co-occurrence networks and influenza gene regulatory networks.