Volatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers

Volatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Volatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers Ting Liu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9015347/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study examines the effectiveness of machine learning and deep learning models in forecasting daily realized volatility and detecting early signs of mar- ket stress. While Transformer-based models have gained popularity for their ability to model long-range dependencies, their performance advantage over clas- sical methods in daily financial settings remain uncertain. Using a 10-year panel of large-cap U.S. equities and a rolling walk-forward evaluation frame- work, we compare traditional time-series models (EWMA, HAR, ARIMA), tree-based ensembles (Random Forest, XGBoost, LightGBM), and deep sequence models (LSTM and Transformer). We assess each model’s forecasting accuracy (RMSE/MAE), rank agreement with realized volatility (Spearman correlation), and ability to detect high-volatility stress episodes using a Market Stress Index (MSI). Results show that simpler models—especially HAR, persistence, and tree ensembles—outperform deep learning approaches in both predictive accuracy and early-warning classification. Although LSTM and Transformer models yield smoother volatility signals, they exhibit weaker rank alignment and lower detec- tion performance at daily frequencies. We also explore how model-implied MSI signals translate into risk-adjusted performance when used in a volatility-targeted overlay strategy. The findings suggest that for daily forecasts with limited input features, classical and ensemble models remain highly effective, while deep architectures may require richer or higher-frequency data to consistently outperform. Financial Mathematics realized volatility forecasting market stress detection market stress index tree ensembles Transformer LSTM volatility-targeted trading walk-forward evaluation Full Text Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9015347","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":599682094,"identity":"1b740b99-1461-4ba7-948b-14214d472820","order_by":0,"name":"Ting Liu","email":"data:image/png;base64,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","orcid":"","institution":"Northwestern University","correspondingAuthor":true,"prefix":"","firstName":"Ting","middleName":"","lastName":"Liu","suffix":""}],"badges":[],"createdAt":"2026-03-03 03:29:17","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":true,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":true},"doi":"10.21203/rs.3.rs-9015347/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9015347/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104401450,"identity":"0728001d-51ba-499a-90f5-5166885c6c48","added_by":"auto","created_at":"2026-03-11 12:12:44","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1352251,"visible":true,"origin":"","legend":"","description":"","filename":"VolatilityForecastingandEarlyWarningMarketStressDetectionALeakageSafeEvaluationwithTreeEnsemblesandTransformers.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9015347/v1_covered_68cdf9c0-b82e-4d72-b958-7520ed460e45.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eVolatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Northwestern University","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"realized volatility forecasting, market stress detection, market stress index, tree ensembles, Transformer, LSTM, volatility-targeted trading, walk-forward evaluation","lastPublishedDoi":"10.21203/rs.3.rs-9015347/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9015347/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study examines the effectiveness of machine learning and deep learning models in forecasting daily realized volatility and detecting early signs of mar- ket stress. While Transformer-based models have gained popularity for their ability to model long-range dependencies, their performance advantage over clas- sical methods in daily financial settings remain uncertain. Using a 10-year panel of large-cap U.S. equities and a rolling walk-forward evaluation frame- work, we compare traditional time-series models (EWMA, HAR, ARIMA), tree-based ensembles (Random Forest, XGBoost, LightGBM), and deep sequence models (LSTM and Transformer). We assess each model’s forecasting accuracy (RMSE/MAE), rank agreement with realized volatility (Spearman correlation), and ability to detect high-volatility stress episodes using a Market Stress Index (MSI). Results show that simpler models—especially HAR, persistence, and tree ensembles—outperform deep learning approaches in both predictive accuracy and early-warning classification. Although LSTM and Transformer models yield smoother volatility signals, they exhibit weaker rank alignment and lower detec- tion performance at daily frequencies. We also explore how model-implied MSI signals translate into risk-adjusted performance when used in a volatility-targeted overlay strategy. The findings suggest that for daily forecasts with limited input features, classical and ensemble models remain highly effective, while deep architectures may require richer or higher-frequency data to consistently outperform.\u003c/p\u003e","manuscriptTitle":"Volatility Forecasting and Early-Warning Market Stress Detection: A Leakage-Safe Evaluation with Tree Ensembles and Transformers","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-04 05:01:46","doi":"10.21203/rs.3.rs-9015347/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e204c412-934b-4e97-a292-dde90604a4a3","owner":[],"postedDate":"March 4th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":63884403,"name":"Financial Mathematics"}],"tags":[],"updatedAt":"2026-03-04T05:01:46+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-04 05:01:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9015347","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9015347","identity":"rs-9015347","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}