Concept-Oriented Graph-Coupled Dynamics for Forecasting

Concept-Oriented Graph-Coupled Dynamics for Forecasting | 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 Article Concept-Oriented Graph-Coupled Dynamics for Forecasting Junbin Gao, Hongwei Ma, Dai Shi, Minh-Ngoc Tran This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9264919/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract Accurate long-horizon forecasting in time series requires more than raw predictive capacity: it demands models that can capture cross-series dependence, preserve multi-scale temporal structure and remain interpretable enough to reveal why a forecast is made. Existing approaches often excel at only part of this objective. Highly expressive deep models typically rely on opaque latent interactions, whereas more interpretable formulations often lack the flexibility required for heterogeneous real-world dynamics. To address this gap, we propose CONCORD, a concept-oriented graph-coupled dynamical forecaster that unifies semantic abstraction, relational reasoning and structured temporal evolution within a single state-space framework. CONCORD first infers explicit multi-scale concept states from causal history, then refines these states through correlation-induced graph coupling, and finally rolls them forward using residual-consistent graph-coupled dynamics. All nonlinear mappings are parameterized by Kolmogorov--Arnold networks, enabling expressive yet functionally transparent representations. This design turns concepts from auxiliary supervision into the core currency through which encoding, interaction and prediction are jointly organized. Across long-term forecasting, short-term traffic forecasting and imputation benchmarks, CONCORD delivers consistently strong performance against competitive recent baselines while providing substantially richer interpretability. Additional analyses show that its learned concept states are stable, semantically structured and predictive of performance-critical temporal patterns across scales. These findings suggest that accurate forecasting and mechanistic interpretability need not be competing goals, and that multivariate temporal prediction can be more effectively approached as the evolution of an explicit, graph-coupled concept state. Physical sciences/Mathematics and computing/Applied mathematics Business and commerce/Information systems and information technology Physical sciences/Mathematics and computing/Computer science Physical sciences/Mathematics and computing/Statistics Full Text Additional Declarations There is NO Competing Interest. Cite Share Download PDF Status: Under Review 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-9264919","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":631487930,"identity":"af8010b4-7bd9-4308-867f-ec5d80886d23","order_by":0,"name":"Junbin Gao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABC0lEQVRIiWNgGAWjYBACxgYGNoYEMJP5AAOQDQYSRGphSyBOCwNcGQOPAXFamGfkHnvwcAeDXX97z+fPPGV35AwOMB+8zcNgl9iAy2Ez8tINEs8wJM84c3abNM+5Z8YGB9iSrXkYkvFoyTGTSGxjSGa4kbuNmbftcOKGAzxm0jwMzIS1yN/IefwZqKV+wwH+b0At9QS12BncyGGQBmpJMDjAwwbUchi3lp435gaJbRIJhmeOmUnOOXfYcOZhNmPLOQbHjXFpMWzPMXv4s83GXu548+MPb8oOy/Mdb354401FtSxOLRAJCbAzmHiAhMJhkIABDvVAIA+l7cGu/AESwWX6KBgFo2AUjFgAAB0uWOP7kx6+AAAAAElFTkSuQmCC","orcid":"","institution":"The University of Sydney","correspondingAuthor":true,"prefix":"","firstName":"Junbin","middleName":"","lastName":"Gao","suffix":""},{"id":631487931,"identity":"222576a5-d35a-402c-9274-7f4849a029e2","order_by":1,"name":"Hongwei Ma","email":"","orcid":"https://orcid.org/0009-0009-2194-8355","institution":"The University of Sydney","correspondingAuthor":false,"prefix":"","firstName":"Hongwei","middleName":"","lastName":"Ma","suffix":""},{"id":631487932,"identity":"781e598f-2fef-4af0-8c9f-aff5bdc75ffc","order_by":2,"name":"Dai Shi","email":"","orcid":"","institution":"The University of Cambridge","correspondingAuthor":false,"prefix":"","firstName":"Dai","middleName":"","lastName":"Shi","suffix":""},{"id":631487933,"identity":"270a590b-548c-4e74-af3a-fb83d130e67d","order_by":3,"name":"Minh-Ngoc Tran","email":"","orcid":"","institution":"The University of Sydney","correspondingAuthor":false,"prefix":"","firstName":"Minh-Ngoc","middleName":"","lastName":"Tran","suffix":""}],"badges":[],"createdAt":"2026-03-30 09:34:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9264919/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9264919/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":109068093,"identity":"28e943f8-e815-42bc-9ac7-3ee4290fd972","added_by":"auto","created_at":"2026-05-12 10:03:35","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1063629,"visible":true,"origin":"","legend":"Article File","description":"","filename":"CONCORDNMI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9264919/v1_covered_32194fa3-bb88-4cb4-83f4-429d779a85dd.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Concept-Oriented Graph-Coupled Dynamics for Forecasting","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-9264919/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9264919/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Accurate long-horizon forecasting in time series requires more than raw predictive capacity: it demands models that can capture cross-series dependence, preserve multi-scale temporal structure and remain interpretable enough to reveal why a forecast is made. Existing approaches often excel at only part of this objective. Highly expressive deep models typically rely on opaque latent interactions, whereas more interpretable formulations often lack the flexibility required for heterogeneous real-world dynamics. To address this gap, we propose CONCORD, a concept-oriented graph-coupled dynamical forecaster that unifies semantic abstraction, relational reasoning and structured temporal evolution within a single state-space framework. CONCORD first infers explicit multi-scale concept states from causal history, then refines these states through correlation-induced graph coupling, and finally rolls them forward using residual-consistent graph-coupled dynamics. All nonlinear mappings are parameterized by Kolmogorov--Arnold networks, enabling expressive yet functionally transparent representations. This design turns concepts from auxiliary supervision into the core currency through which encoding, interaction and prediction are jointly organized. Across long-term forecasting, short-term traffic forecasting and imputation benchmarks, CONCORD delivers consistently strong performance against competitive recent baselines while providing substantially richer interpretability. Additional analyses show that its learned concept states are stable, semantically structured and predictive of performance-critical temporal patterns across scales. These findings suggest that accurate forecasting and mechanistic interpretability need not be competing goals, and that multivariate temporal prediction can be more effectively approached as the evolution of an explicit, graph-coupled concept state.","manuscriptTitle":"Concept-Oriented Graph-Coupled Dynamics for Forecasting","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-12 06:47:56","doi":"10.21203/rs.3.rs-9264919/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"9b1d7c92-3a45-4321-a2be-5c597ee85640","owner":[],"postedDate":"May 12th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":67212004,"name":"Physical sciences/Mathematics and computing/Applied mathematics"},{"id":67212005,"name":"Business and commerce/Information systems and information technology"},{"id":67212006,"name":"Physical sciences/Mathematics and computing/Computer science"},{"id":67212007,"name":"Physical sciences/Mathematics and computing/Statistics"}],"tags":[],"updatedAt":"2026-05-12T06:47:56+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-12 06:47:56","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9264919","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9264919","identity":"rs-9264919","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}