Reproduce “Institutional Equity Holdings Prediction Using Node Affinities of Dynamic Graphs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Emad Izadifar, Zahed Rahmati
This research establishes a benchmark for predicting institutional equity holdings by formulating it as a dynamic node affinity prediction task on a temporal bipartite graph, leveraging SEC Form 13F data. The study demonstrates that the NAVIS model achieves the highest prediction accuracy (NDCG@10 of 0.9127), outperforming other temporal graph neural networks, while simple heuristic baselines also show strong performance due to the persistence of institutional portfolios.
Key insights
- Institutional portfolio allocations exhibit significant temporal persistence, making simple heuristic methods like Persistent Forecast and Exponential Moving Average highly competitive.
- Temporal graph representation learning, specifically with models designed for node affinity, can effectively capture complex dynamics in institutional holdings, surpassing other TGNNs.
- Explicit domain-specific node features provide only marginal performance improvements, suggesting that the underlying temporal and structural signals within the ownership graph are highly informative.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.