{"paper_id":"2401.06249","title":"SpotV2Net: Multivariate Intraday Spot Volatility Forecasting via Vol-of-Vol-Informed Graph Attention Networks","authors":["Alessio Brini","Giacomo Toscano"],"paper_date":"2024-01-11","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2401.06249v4","page_url":"https://thequant.space/flowcharts/spotv2net-multivariate-intraday-spot-volatility-forecasting/","scores":{"empirical_rigor":8,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes SpotV2Net, a Graph Attention Network model for multivariate intraday spot volatility forecasting that uses Fourier‑based spot volatility and volatility‑of‑volatility features as node and edge attributes.","quote":"This paper introduces SpotV2Net, a multivariate intraday spot volatility forecasting model based on a Graph Attention Network architecture.","summary":"This paper introduces SpotV2Net, a multivariate intraday spot volatility forecasting model based on a Graph Attention Network architecture. SpotV2Net represents assets as nodes within a graph and includes non-parametric high-frequency Fourier estimates of the spot volatility and co-volatility as…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The log‑price and spot volatility processes follow diffusion dynamics with continuous, bounded, adapted drift and volatility‑of‑volatility processes.","type":"model","stated":"explicit","quote":"... the drift µi,k and the spot volatility of volatility eVi are continuous, bounded and adapted stochastic processes"},{"assumption":"The graph of assets is fully connected (all adjacency entries equal to 1).","type":"model","stated":"explicit","quote":"The graph G is assumed to be fully connected, that is, all the entries of A are assumed to be equal to 1."},{"assumption":"High‑frequency tick data are filtered to NYSE trades within regular market hours and resampled to one‑second intervals.","type":"data","stated":"explicit","quote":"we filter the transactions to include only those executed on the New York Stock Exchange (NYSE) and exclude transactions that occurred outside the operating market hours from 9:30 a.m. to 4:00"},{"assumption":"Two hidden GAT layers are optimal for the SpotV2Net architecture.","type":"model","stated":"explicit","quote":"When implementing SpotV2Net, we optimize the number of hidden layers as a hyperparameter, determining that two hidden layers are optimal."}]},"datasets":{"status":"available","items":[{"name":"DJIA component high‑frequency price data","provider":"TAQ – Millisecond Consolidated Trades database accessed via ","kind":"market-prices","access":"commercial","period":"June 1st, 2020 to May 10th, 2023","universe":"30 components of the Dow Jones Industrial Average","quote":"We collect the raw price data for the 30 components of the DJIA index from the TAQ - Millisecond Consolidated Trades database, which we access through the Wharton Research Data Services (WRDS). The dataset covers the period from June 1st, 2020, to May 10th, 2023, corresponding to 737 days.","known_id":"taq"}],"known":[{"id":"the-graph","name":"The Graph (subgraphs)","provider":"The Graph","kind":"crypto-onchain","access":"public","quote":"Note that we use GNNEXPLAINER to identify the most influential subgraph, not to isolate the most influential individual features of specific nodes.","mentions":19,"used_in":5},{"id":"wrds","name":"WRDS","provider":"Wharton Research Data Services","kind":"other","access":"commercial","quote":"7.1 Data preparation and estimation of univariate and multivariate spot volatility and the volatility of volatility series We collect the raw price data for the 30 components of the DJIA index from the TAQ - Millisecond Consolidated Trades database, which we access through the Wharton Research Data Services (WRDS).","mentions":2,"used_in":1},{"id":"taq","name":"NYSE TAQ","provider":"NYSE","kind":"market-prices","access":"commercial","quote":"7.1 Data preparation and estimation of univariate and multivariate spot volatility and the volatility of volatility series We collect the raw price data for the 30 components of the DJIA index from the TAQ - Millisecond Consolidated Trades database, which we access through the Wharton Research Data Services (WRDS).","mentions":1,"used_in":1}],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"tick"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"n/a","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"SpotV2Net achieves lower MSE and QLIKE than HAR‑Spot, XGB, and LSTM in both validation and test periods, with statistical significance.","kind":"predictive-accuracy","quote":"Aggregate performance measures in Tab. 2 suggest that SpotV2Net outperforms the HAR‑Spot model and the other ML models considered, namely XGB and LSTM, both during the validation and test periods."},{"finding":"Model Confidence Set analysis includes only SpotV2Net at the 95% confidence level, indicating its superior predictive performance.","kind":"statistical","quote":"models contains only SpotV2Net, indicating that the latter yields superior predictive performance, compared to the other models considered, with a level of confidence of 95%."},{"finding":"Diebold‑Mariano tests confirm that the forecasting accuracy gains of SpotV2Net over competing models are statistically significant at the 5% level.","kind":"statistical","quote":"The results of the DM test suggest that the gain in forecasting accuracy provided by SpotV2Net is statistically significant."},{"finding":"Edge features (spot co‑volatility of volatility) are crucial; removing them (SpotV2Net‑NE) reduces accuracy, confirming the importance of spillover modeling.","kind":"performance","quote":"These findings underline the crucial role of edge features (that is, the values of the spot co‑volatility of volatility matrix) in capturing spillover effects and improving forecasting accuracy."},{"finding":"Interpretability analysis shows that Amgen, Salesforce, and Honeywell are the most influential nodes for predicting intraday spot volatility across all DJIA stocks.","kind":"other","quote":"Fig. 4 shows that Amgen (AMGN), Salesforce (CRM) and Honeywell (HON) are deemed the three most influential nodes for predicting the intraday spot volatility of all stocks in the dataset, in both the validation and test phases."}],"reported_performance":null,"out_of_sample":{"method":"holdout","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":40,"citations":{"cited_by_in_corpus":0,"by_independent_authors":0,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:16:27Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}