Multi-Teacher Knowledge Distillation for LightweightFood Price Forecasting

Multi-Teacher Knowledge Distillation for LightweightFood Price 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 Multi-Teacher Knowledge Distillation for LightweightFood Price Forecasting Shifat Zaman, Md. Golam Rabiul Alam This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8699707/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 Food price volatility poses significant challenges to humanitarian organizations operating in resource-constrained environments.While deep learning models achieve state-of-the-art forecasting accuracy, their computational requirements limit deploymentin field settings. We present a multi-teacher knowledge distillation framework that transfers ensemble knowledge from threearchitecturally diverse teacher models—DLinear, PatchTST, and N-BEATS—to a lightweight student network. Our frameworkintroduces a four-component distillation loss combining hard loss, uncertainty-weighted prediction distillation, feature distillation,and difference distillation for time series regression. Evaluated on World Food Programme Bangladesh price data for fouressential commodities, the framework achieves 1.959 BDT Mean Absolute Error and 3.73% Mean Absolute Percentage Error,representing 37% improvement over supervised baselines. The 200K-parameter student model enables CPU-based inferencewhile retaining 97% of teacher ensemble accuracy, providing a practical solution for humanitarian food security applications. Physical sciences/Engineering Physical sciences/Mathematics and computing Full Text Additional Declarations No competing interests reported. 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-8699707","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":583495430,"identity":"3da26d68-1514-429a-95ad-a034576d5863","order_by":0,"name":"Shifat Zaman","email":"data:image/png;base64,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","orcid":"","institution":"BRAC University","correspondingAuthor":true,"prefix":"","firstName":"Shifat","middleName":"","lastName":"Zaman","suffix":""},{"id":583495431,"identity":"ba250683-2458-4867-9337-a04ef67a387b","order_by":1,"name":"Md. Golam Rabiul Alam","email":"","orcid":"","institution":"BRAC University","correspondingAuthor":false,"prefix":"","firstName":"Md.","middleName":"Golam Rabiul","lastName":"Alam","suffix":""}],"badges":[],"createdAt":"2026-01-26 11:24:51","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8699707/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8699707/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107705426,"identity":"e2153b3d-f94f-4bc8-9cf7-11bab4dfef79","added_by":"auto","created_at":"2026-04-24 09:12:33","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":346315,"visible":true,"origin":"","legend":"","description":"","filename":"RevisedSubmission.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8699707/v1_covered_a04db439-8bb7-41e6-b58d-7e69a6b98d02.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multi-Teacher Knowledge Distillation for LightweightFood Price Forecasting","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"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":"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":"","lastPublishedDoi":"10.21203/rs.3.rs-8699707/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8699707/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Food price volatility poses significant challenges to humanitarian organizations operating in resource-constrained environments.While deep learning models achieve state-of-the-art forecasting accuracy, their computational requirements limit deploymentin field settings. We present a multi-teacher knowledge distillation framework that transfers ensemble knowledge from threearchitecturally diverse teacher models—DLinear, PatchTST, and N-BEATS—to a lightweight student network. Our frameworkintroduces a four-component distillation loss combining hard loss, uncertainty-weighted prediction distillation, feature distillation,and difference distillation for time series regression. Evaluated on World Food Programme Bangladesh price data for fouressential commodities, the framework achieves 1.959 BDT Mean Absolute Error and 3.73% Mean Absolute Percentage Error,representing 37% improvement over supervised baselines. The 200K-parameter student model enables CPU-based inferencewhile retaining 97% of teacher ensemble accuracy, providing a practical solution for humanitarian food security applications.","manuscriptTitle":"Multi-Teacher Knowledge Distillation for LightweightFood Price Forecasting","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-05 07:21:22","doi":"10.21203/rs.3.rs-8699707/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":"3601c486-df51-4832-895b-a049e9e06a47","owner":[],"postedDate":"February 5th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62082035,"name":"Physical sciences/Engineering"},{"id":62082036,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-04-22T09:13:19+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-05 07:21:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8699707","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8699707","identity":"rs-8699707","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}