{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T14:40:28Z","timestamp":1792248028852,"version":"4.4.0"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":0,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/www.uspto.gov\/ip-policy\/copyright-policy\/copyright-basics"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>Interbank networks transmit liquidity and systemic risk, making forecasts of their topology,edge weights, and node values important for forward-looking monitoring of risk concentration and contagion. Joint forecasting is difficult under structural breaks and macro-financialregime changes. We propose the Frequency-Disentangled Graph Mixture-of-Experts (FDGMoE), which combines a dual-branch graph encoder with regime-aware adaptive gating. Theencoder separates slow-varying structural patterns from high-frequency local interactions, while three temporal experts capture long-term, cyclical, and transient dynamics. A joint prediction module forecasts network topology, edge weights, and node values. We evaluate FDG-MoE on weighted directed networks reconstructed from annual balance-sheet data for 70 Chinese banks from 2007 to 2023, using 2007\u20132019 for training and 2020\u20132023 for testing. Relative to the strongest baseline for each metric, FDG-MoE increases weight similarity by 0.30% and reduces Kullback\u2013Leibler divergence between predicted and target degree distributions by 8.83%. It also reduces mean squared error and mean absolute error for node values by 21.37% and 19.86%, respectively. Ablation and case results further demonstrate the effectiveness of its components. From an artificial intelligence perspective, the contribution is a frequency-disentangled graph MoE architecture that integrates slow\u2013fast graph representation learning, regime-aware expert routing, and joint multi-target prediction for non-stationary graph sequences. From an engineering application perspective, the contribution is a validated interbank forecasting workflow that jointly estimates future counterparty links, exposure weights, and bank-level financial values from balance-sheet-derived networks, providing structured inputs for forward-looking monitoring of risk concentration and contagion<\/jats:p>","DOI":"10.2139\/ssrn.7624658","type":"posted-content","created":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T13:45:06Z","timestamp":1792244706000},"source":"Crossref","is-referenced-by-count":0,"title":["FDG-MoE: Frequency-Disentangled Graph Mixture-of-Experts for Non-Stationary Interbank Network Forecasting"],"prefix":"10.2139","author":[{"given":"Huanlan","family":"Yan","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-9134-3484","authenticated-orcid":true,"given":"binbin","family":"feng","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhijun","family":"Ding","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,10,17]],"date-time":"2026-10-17T13:45:06Z","timestamp":1792244706000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7624658"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":0,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7624658","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}