This study investigates how globalization has reshaped the world academic migration network. Using affiliation changes across almost 50 million Scopus articles, I construct five-decade-aggregated networks covering the period from 1974 to 2023. Three findings stand out. The network densifies and grows more equal: more country pairs are connected in each successive decade, and the share of flows captured by the largest hubs falls steadily, even as the United States retains the central position throughout. Reciprocity stays high at every threshold and in every decade, so academic migration is circulatory in structure rather than a one-directional drain from periphery to core. The rise of Asia takes the form of integration of China, India, and South Korea into the US-centered core, not a separate Asian cluster.
This study proposes a reduced-form, access-based model of network growth in which attachment depends jointly on accumulated connectivity, residual capacity, and origin–destination pair-specific link cost. The kernel contains Barabási–Albert (BA) preferential attachment, spatial attachment, and capacity-limited growth as limiting cases. At N=1000, α=1, and κ=2, the capacity-only specification ( β=1 , K=16 ) reduces mean maximum degree from 82.31 in the BA limiting case to 16.00, while the spatial-only specification ( ϕ=1 ) reduces mean edge length from 0.521 to 0.351 . Increasing ϕ from 0 to 2 shortens mean edge length from 0.526 to 0.157, and increasing κ from 1 to 4 raises clustering and reduces average path length. With mean capacity near 16 , integer-uniform and rounded-lognormal capacities raise mean maximum degree to 23.06 and 49.00 , respectively. Crossing-handling and access-guided extensions represent junction formation and compare four configured transport-growth strategies using a common ex post access measure. The framework yields directional hypotheses for empirical calibration.
This study examines the mesoscale organization of Indonesia’s intranational production network using the 2016 Interregional Input–Output (IRIO) table. We represent intermediate transactions as a weighted, directed supra-network of 578 province–sector nodes and combine three network tools in a single pipeline, namely consensus Leiden community detection, weighted rich-club analysis, rich-node enrichment testing, and inter-community backbone extraction. We evaluate whether production communities are primarily organized by geography or sectoral alignment, assess how economically central province–sector nodes are distributed across communities, and examine how the inter-community backbone structures national production integration. Results show that communities are predominantly geographic, aligning strongly with provinces and island groupings rather than sector labels, and therefore form spatially cohesive but multi-sector production blocs. The sector groups that dominate internal value added and external linkage channels further characterize the production-network profile of each community. This comparison identifies Java-based manufacturing production cores, Jakarta’s service-intermediation role, East Java’s hybrid service–manufacturing role, and resource-processing or resource-supplying interfaces outside Java. Rich-club and enrichment analyses show that production-network centrality is organized as a pattern of multiple spatially grounded rich-core communities rather than being absorbed into a single dominant production community. This multiple rich-core structure is Java-centered at the national mesoscale, with the island-constrained test giving the strongest support to the Jakarta-centered community. Cross-resolution results further show that selected communities outside Java, especially in resource-producing regions, exhibit supplier-side rich-node concentrations at finer spatial resolution. The inter-community backbone reveals a Java-centered strongest-corridor structure linking western, central, and eastern Indonesia through Jakarta and East Java. A redundancy-preserving PMFG comparison shows that these two communities remain structurally important gateway positions when additional high-weight routes are retained. Based on the 2016 IRIO table, the findings offer a structural snapshot of flow concentration and corridor dependence, serving as a basis for future dynamic analysis of shock propagation. Overall, the mesoscale framework provides a systematic basis for understanding how spatial production communities, rich-node concentration, and interregional gateway corridors jointly shape Indonesia’s production network, with implications for spatial inequality, national integration, and structural dependence.
Information propagation over graph-structured data depends critically on the quality of node representations learned by graph neural networks (GNNs). However, how different GNN architectures organize embedding spaces and how these representations influence downstream propagation behavior remain insufficiently understood. This study presents a comprehensive empirical evaluation of ten representative GNN architectures on six attributed graph datasets using a unified unsupervised autoencoding framework and a fixed similarity-based propagation protocol. Rather than simulating a specific real-world diffusion process, the proposed framework provides a controlled setting for comparing how learned embeddings support propagation under identical conditions. Propagation behavior is quantified using normalized cascade size and multi-step activation dynamics. Temporal propagation patterns are analyzed using Dynamic Time Warping (DTW), while embedding organization is examined through principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), and clustering analysis. To investigate the role of graph topology, thirteen structural network metrics are analyzed using Pearson correlation, mixed-effects modeling, and SHapley Additive exPlanations (SHAP), followed by controlled graph intervention experiments. The results show that propagation behavior is jointly influenced by graph topology and the representations learned by different GNN architectures. PNA, GAT, GCN, and EvolveGCN consistently produce embedding spaces associated with larger propagation cascades across multiple datasets, although no single architecture is uniformly superior. The analyses further reveal architecture-specific structural fingerprints and demonstrate that identical topological perturbations can induce markedly different propagation responses. These findings provide new insights into how GNN representation learning shapes downstream similarity-based propagation behavior. The data and codes used in this study are available to reviewers at https://doi.org/10.5281/zenodo.17204159. They will be made publicly available upon acceptance of the manuscript.
The COVID-19 pandemic accelerated the development of artificial intelligence (AI), machine learning (ML), and deep learning (DL) for clinical decision support, public-health surveillance, and pandemic response. This systematic review evaluates AI-based studies addressing COVID-19 diagnosis, prognosis, outbreak forecasting, remote monitoring, digital epidemic intelligence, and multimodal applications, emphasizing dataset characteristics, validation maturity, generalizability, fairness, privacy, and deployment readiness. Following PRISMA 2020, studies published between December 2019 and 31 January 2024 were retrieved from IEEE Xplore, ACM Digital Library, SpringerLink, ScienceDirect, Wiley Online Library, and arXiv using database-adapted search strategies. Of 110 identified records, 57 met the eligibility criteria and underwent a six-domain CASP-informed methodological appraisal and validation-maturity assessment. Quantitative synthesis characterized application domain, publication timing, dataset geographic provenance and scale, principal model family, imaging modality, and validation strategy. Diagnostic imaging and clinical diagnosis constituted the largest evidence domain, followed by prognosis, forecasting, remote monitoring/IoT, OSN/NLP surveillance, and multimodal or hybrid systems. Most studies remained at the development-validation stage: 46 of 57 (80.7
