《Sentiment Predictability for Stocks》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。

1、SentimentPredictabilityforStocksJordanProsky1,AndrewTan2,MichaelZhao3,XingyouSong4Abstract—Inthiswork,wepresentour?ndingsandex-VIConclusions8perimentsforstock-marketpredictionusingvarioustextualsentimentanalysistools,suchasmoodanalysisandeventVIIAPPENDIX9extraction,aswell

2、aspredictionmodels,suchasLSTMsandspeci?cconvolutionalarchitectures.References9CONTENTSI.INTRODUCTIONIIntroduction1Stockmarketpredictionisanimportanttime-serieslearn-ingproblemin?nanceandeconomics.AccordingtotheIIRelevantandPreviousWorks1semi-strongformoftheef?cientmarketh

3、ypothesis(EMH),II-AAggregateTwitterPredictionofDowmarketsareef?cientandstockpricesalreadyre?ectallpub-Jones..................2liclyavailableinformationsuchas?nancialnewsandpriceII-BStockMovementPredictionUsingdata.Thishypothesisincorporatestheweak-formEMH,NewsArticleTitle

4、swithSVMs...2whichimpliesthatstockpricesre?ectallmarketinformation,II-CDeepLearninginFinance.......2asubsetofpublicdata.Semi-strongformEMHassumesIIIIndividualStockPrediction2thatthemarketadjustsquicklytoabsorbnewinformationIII-ADif?culties...............2andstockpricesre?

5、ectallnewavailableinformation.GivenIII-BAdvancedTextualInformationAp-thisassumptionandthefactthatinvestorspurchasestocksproaches................3afterpublicinformationisreleased,aninvestortheoreticallycannotbene?toroutperformthemarketbytradingonnewIVExperiments3informatio

6、n.Severalstudiescontradictthishypothesis,andIV-AStockPriceandFinancialNewsData3throughourresearch,weshowthatweareabletooutperformIV-A.1WebsitesUsed.......3themarketusingacombinationofdeeplearning,temporalIV-A.2SentimentAnalyzer....3data,andnewsarticles.IV-A.3OpenIEExtract

7、orandOurreportfocusesontheuseofvarioustime-seriesandWord2Vec.........3NLPtechniquestoextractinformationaboutanindividualIV-A.4DeepLearningPackages.3stock’schanceofclosinghigherorlowerthanitsopeningIV-BSetup..................3price.WetakeadvantageofthetemporalnatureofbothI

8、V-B.1LSTM-BasedMoodPre-stockpricedataand?nancialnewsdata.InSection2,wediction...........3explore