Forecasting the Short-term Price Trend of Taiwan Stocks

ForecastingtheShort-termPriceTrendofTaiwanStockswithDeepNeuralNetworkMing-CheLeeDepartmentofComputerandCommunicationEngineering,MingChuanUniversityNo.5DeMingRd.,GuiShanDistrict,TaoyuanCounty333,Taiwan+883-3-3507001#3320leemc@mail.mcu.edu.twJie-ShanLiaoDepartmentofComputerandCommunicationEngineering,MingChuanUniversityNo.5DeMingRd.,GuiShanDistrict,TaoyuanCounty333,Taiwan+883-3-3507001#3320s9513369@gmail.comSheng-ChengYehDepartmentofComputerandCommunicationEngineering,MingChuanUniversityNo.5DeMingRd.,GuiShanDistrict,TaoyuanCounty333,Taiwan+883-3-3507001#3320peteryeh@mail.mcu.edu.twJia-WeiChangDepartmentofComputerScienceandInformationEngineering,NationalTaichungUniversityofScienceandTechnologyNo.129,Sec.3,SanminRd.,NorthDist.,TaichungCity40401,Taiwan+886-4-22196320jwchang@nutc.edu.twABSTRACTTherelationshipbetweentechnicalindicatorsandpricesinstockmarkethasalwaysbeenanimportanttopicofconcernfortheacademicandfinancialcommunities.Manyliteraturessuggestthatitisfeasibletousetechnicalanalysistoestimatethefuturepriceofstocks.Theuseofmachinelearningtoestimatestockpriceshasalsograduallybecomemainstreaminthefinancialmarket.Thisstudyaimstoexplorethefeasibilityofusingdeepnetworkandtechnicalanalysisindicatorstoestimateshort-termpricemovementsofstocks.ThesubjectofthisstudyisTWSE0050,whichisthemosttradedETFinTaiwan'sstockexchange.WeuseLongShortTermMemory(LSTM)toconstructadeepnetworkstockestimationmodelandconductexperimentsontheTaiwanStockExchange'sopendatafrom2019/01to2019/10.Experimentalresu