Bogumił Kamiński

Efficient communication in many types of dynamic networks depends critically on how nodes are clustered into subnetworks. As such networks grow and evolve rapidly, there is a need for fast and robust clustering algorithms that account for communication cost structures induced by different node partitions. In this paper, we evaluate how classic community detection algorithms (CDAs), i.e. the Louvain and Ensemble Clustering for Graphs (ECG), adapted to incorporate a flexible family of cost functions, perform relative to standard heuristic and metaheuristic approaches. Using simulations on l-nearest neighbour graphs of varying sizes, we find that especially the Louvain algorithm consistently delivers high quality solutions at a substantially lower computational cost. In contrast, metaheuristic methods fail to scale effectively, and the ECG algorithm does not provide performance improvements in our setting, despite its reported stabilising effect in traditional community detection tasks. Overall, our results indicate that classic CDAs are well suited for real time clustering in dynamic communication networks and constitute a strong basis for developing scalable communication optimisation strategies.

Community detection methods play a central role in understanding complex networks by revealing highly connected subsets of entities. However, most community detection algorithms generate partitions of the nodes, thus (i) forcing every node to be part of a community and (ii) ignoring the possibility that some nodes may be part of multiple communities. In our work, we investigate three simple community association strength (CAS) scores and their usefulness as post-processing tools given some partition of the nodes. We show that these measures can be used to improve node partitions, detect outlier nodes (not part of any community), and help find nodes with multiple community memberships.

This paper shows how information about the network’s community structure can be used to define node features with high predictive power for classification tasks. To do so, we define a family of community-aware node features and investigate their properties. Those features are designed to ensure that they can be efficiently computed even for large graphs. We show that community-aware node features contain information that cannot be completely recovered by classical node features or node embeddings (both classical and structural) and bring value in node classification tasks. This is verified for various classification tasks on synthetic and real-life networks.

Drawing upon institutional theory, we investigate how companies react to coercive pressures which impose anti-corruption disclosure practices. We adopt the concept of change in the institutionalized field and investigate the impact of the natural experiment of the Non-Financial Reporting Directive (NFRD) implementation on a company’s choice for disclosing its anti-corruption policy. We examine the relationship between firm linkages with the external environment, proxied by board independence and ownership dispersion, and anti-corruption disclosure. We use a sample of 72 companies listed on the Warsaw Stock Exchange over the period of 2015–2019 that were subject to the NFRD legislation. The evidence from the Tobit model shows that the linkages with the external environment differentiate company reactions to the implementation of the mandatory reporting legislation. In particular, greater company linkages via interdependent directors and ownership dispersion increase the scope of the anti-corruption disclosure in the post-NFRD period. Our study offers policy implications suggesting that corporate scrutiny and exposure to external constituencies may improve implementation of legislation into company practice and enhance anti-corruption disclosure.