基于大數(shù)據(jù)的共享單車預測研究的國內(nèi)外文獻綜述在當前的大數(shù)據(jù)時代中，每一秒鐘都有大量的數(shù)據(jù)產(chǎn)生，大數(shù)據(jù)促進了統(tǒng)計學科的發(fā)展，由數(shù)據(jù)驅(qū)動的統(tǒng)計方法現(xiàn)在被廣泛地應(yīng)用在統(tǒng)計應(yīng)用研究中。李金昌ADDINEN.CITE<EndNote><Cite><Author>李金昌</Author><Year>2015</Year><RecNum>98</RecNum><DisplayText><styleface="superscript">[4]</style></DisplayText><record><rec-number>98</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824506">98</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>李金昌</author></authors></contributors><auth-address>浙江財經(jīng)大學;</auth-address><titles><title>統(tǒng)計測度:統(tǒng)計學邁向數(shù)據(jù)科學的基礎(chǔ)</title><secondary-title>統(tǒng)計研究</secondary-title></titles><periodical><full-title>統(tǒng)計研究</full-title></periodical><pages>3-9</pages><volume>32</volume><number>08</number><keywords><keyword>統(tǒng)計測度</keyword><keyword>統(tǒng)計學</keyword><keyword>大數(shù)據(jù)</keyword><keyword>數(shù)據(jù)科學</keyword></keywords><dates><year>2015</year></dates><isbn>1002-4565</isbn><call-num>11-1302/C</call-num><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[4]指出統(tǒng)計測度就是用符號和數(shù)字等形式和載體，對所研究的事物的特征進行量化反映，統(tǒng)計思維、統(tǒng)計方法與計算技術(shù)相結(jié)合的基礎(chǔ)就是科學的統(tǒng)計測度。為了預測共享單車站點的需求，我們必須從數(shù)據(jù)的內(nèi)在機理出發(fā)，考慮影響人們使用共享單車出行的因素，以及人們使用共享單車的出行規(guī)律，將人們的出行規(guī)律和數(shù)據(jù)挖掘技術(shù)結(jié)合起來，在科學有效的統(tǒng)計測度下進行預測模型的構(gòu)建。伴隨著共享單車系統(tǒng)的快速發(fā)展，國內(nèi)外學者關(guān)于這一交通方式的研究也不斷深入，主要針對公共自行車發(fā)展過程中呈現(xiàn)的需求預測及調(diào)度優(yōu)化問題進行了探索。同時，共享單車出行量預測屬于交通流量管理類問題，其他交通流量類問題的研究思路尚可幫助到此問題的研究。針對影響共享單車需求的因素，徐長興等建立了基于格蘭杰因果分析和相似日選擇的組合預測模型，在對北京市2017年5月10日-31日的共享單車出行數(shù)據(jù)的分析中，從刻畫天氣因素的指標中選取了溫度、風速、濕度和氣壓4個天氣指標作為影響共享單車需求的天氣因素，同時基于灰色關(guān)聯(lián)分析來選擇相似日作為共享單車需求量預測的歷史數(shù)據(jù)。機器學習算法能夠綜合考慮出行需求的時間序列特征和外部影響因素。陳星佑ADDINEN.CITE<EndNote><Cite><Author>陳星佑</Author><Year>2020</Year><RecNum>96</RecNum><DisplayText><styleface="superscript">[5]</style></DisplayText><record><rec-number>96</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824463">96</key></foreign-keys><ref-typename="Thesis">32</ref-type><contributors><authors><author>陳星佑</author></authors><tertiary-authors><author>張立文,</author></tertiary-authors></contributors><titles><title>基于分層聚類及LSTM模型的共享單車流量預測研究</title></titles><dates><year>2020</year></dates><publisher>上海財經(jīng)大學</publisher><work-type>碩士</work-type><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[5]在對共享單車站點需求量進行預測前對出行數(shù)據(jù)進行了聚類，考慮了站點的地理位置和共享單車在站點間的流量轉(zhuǎn)換。華明壯ADDINEN.CITE<EndNote><Cite><Author>華明壯</Author><Year>2018</Year><RecNum>88</RecNum><DisplayText><styleface="superscript">[6]</style></DisplayText><record><rec-number>88</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824448">88</key></foreign-keys><ref-typename="Thesis">32</ref-type><contributors><authors><author>華明壯</author></authors><tertiary-authors><author>陳學武,</author></tertiary-authors></contributors><titles><title>基于訂單數(shù)據(jù)挖掘的共享單車調(diào)度需求分析方法研究</title></titles><keywords><keyword>共享單車</keyword><keyword>訂單數(shù)據(jù)</keyword><keyword>出行特征</keyword><keyword>聚類分析</keyword><keyword>客流預測</keyword><keyword>調(diào)度需求</keyword></keywords><dates><year>2018</year></dates><publisher>東南大學</publisher><work-type>碩士</work-type><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[6]對共享單車訂單數(shù)據(jù)進行聚類時使用了K-means聚類、DBSCAN聚類和時空聚類，比較后得出DBSCAN聚類效果不如K-means聚類和時空聚類，K-means聚類結(jié)果的輪廓系數(shù)指標更優(yōu)，但時空聚類從人們選擇共享單車出行的邏輯上出發(fā)，考慮了出行需求隨時間變化的因素。研究表明影響單車需求量的外部因素主要包括天氣因素(溫度、降水量、風速等)ADDINEN.CITE<EndNote><Cite><Author>Mattson</Author><Year>2017</Year><RecNum>120</RecNum><DisplayText><styleface="superscript">[7]</style></DisplayText><record><rec-number>120</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">120</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>JeremyMattson</author><author>RanjitGodavarthy</author></authors></contributors><auth-address>SmallUrbanandRuralTransitCenter,UpperGreatPlainsTransportationInstitute,NorthDakotaStateUniversity,NDSUDept.2880,P.O.Box6050,Fargo,ND58108-6050,UnitedStates</auth-address><titles><title>BikeshareinFargo,NorthDakota:Keystosuccessandfactorsaffectingridership</title><secondary-title>SustainableCitiesandSociety</secondary-title></titles><periodical><full-title>SustainableCitiesandSociety</full-title></periodical><volume>34</volume><keywords><keyword>Bikeshare</keyword><keyword>Smallurban</keyword><keyword>Universitytransportation</keyword><keyword>Keystosuccess</keyword><keyword>Ridership</keyword></keywords><dates><year>2017</year></dates><isbn>2210-6707</isbn><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[7]到人口統(tǒng)計特征、建筑環(huán)境特征ADDINEN.CITE<EndNote><Cite><Author>Campbell</Author><Year>2016</Year><RecNum>121</RecNum><DisplayText><styleface="superscript">[8]</style></DisplayText><record><rec-number>121</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">121</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>AndrewA.Campbell</author><author>ChristopherR.Cherry</author><author>MeganS.Ryerson</author><author>XinmiaoYang</author></authors></contributors><auth-address>CivilandEnvironmentalEngineering,UniversityofCalifornia,Berkeley,CA94720,UnitedStates;;CivilandEnvironmentalEngineering,321JDTickleBuilding,TheUniversityofTennessee,Knoxville,TN37996-2313,UnitedStates;;DepartmentofCityandRegionalPlanning,102MeyersonHall,210South34thStreet,UniversityofPennsylvania,Philadelphia,PA19104,UnitedStates;;InstituteofTransportationEngineering,DepartmentofCivilandEnvironmentalEngineering,TsinghuaUniversity,Beijing100084,China</auth-address><titles><title>FactorsinfluencingthechoiceofsharedbicyclesandsharedelectricbikesinBeijing</title><secondary-title>TransportationResearchPartC</secondary-title></titles><periodical><full-title>TransportationResearchPartC</full-title></periodical><volume>67</volume><keywords><keyword>Bikeshare</keyword><keyword>E-bike</keyword><keyword>Statedpreference</keyword><keyword>Bicycle</keyword><keyword>Transit</keyword><keyword>Choicemodeling</keyword></keywords><dates><year>2016</year></dates><isbn>0968-090X</isbn><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[8]和位置因素ADDINEN.CITE<EndNote><Cite><Author>Riexy</Author><Year>2012</Year><RecNum>125</RecNum><DisplayText><styleface="superscript">[9]</style></DisplayText><record><rec-number>125</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621844806">125</key></foreign-keys><ref-typename="Book">6</ref-type><contributors><authors><author>RARiexy</author></authors></contributors><titles><title><styleface="normal"font="default"size="100%">Station-levelforecastingofbikesharingridership:stationnetworkeffectsinthreeU</style><styleface="normal"font="default"charset="134"size="100%">．</style><styleface="normal"font="default"size="100%">S</style><styleface="normal"font="default"charset="134"size="100%">．</style><styleface="normal"font="default"size="100%">Systems</style></title></titles><dates><year>2012</year></dates><pub-location>London</pub-location><publisher>SAGEPublications</publisher><urls></urls></record></Cite></EndNote>[9]和交通事件等ADDINEN.CITE<EndNote><Cite><Author>Chen</Author><Year>2016</Year><RecNum>124</RecNum><DisplayText><styleface="superscript">[10]</style></DisplayText><record><rec-number>124</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">124</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>LongbiaoChen</author><author>DaqingZhang</author><author>LeyeWang</author><author>DingqiYang</author><author>XiaojuanMa</author><author>ShijianLi</author><author>ZhaohuiWu</author><author>GangPan</author><author>Thi-Mai-TrangNguyen</author><author>JérémieJakubowicz</author></authors></contributors><auth-address>ZhejiangUniversity,ChinaandCNRSSAMOVAR,FranceandSorbonneUniversites,UPMCUnivParis06,UMR,LIP6,4PlaceJussieu,Paris,France;;CNRSSAMOVAR,FranceandPekingUniversity,China;;CNRSSAMOVAR,France;;UniversityofFribourg,Switzerland;;HongKongUniversityofScienceandTechnology,HongKong;;ZhejiangUniversity,China;;ZhejiangUniversity,China;;ZhejiangUniversity,China;;SorbonneUniversites,UPMCUnivParis06,UMR,LIP6,4PlaceJussieu,Paris,France;;CNRSSAMOVAR,France</auth-address><titles><title>Dynamiccluster-basedover-demandpredictioninbikesharingsystems</title><secondary-title>PervasiveandUbiquitousComputing</secondary-title></titles><periodical><full-title>PervasiveandUbiquitousComputing</full-title></periodical><keywords><keyword>Activityrecognition</keyword><keyword>Ontologicalreasoning</keyword><keyword>Probabilisticreasoning</keyword></keywords><dates><year>2016</year></dates><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[10]因素的影響。種穎珊等ADDINEN.CITE<EndNote><Cite><Author>種穎珊</Author><Year>2018</Year><RecNum>117</RecNum><DisplayText><styleface="superscript">[11]</style></DisplayText><record><rec-number>117</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">117</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>種穎珊</author><author>韓曉明</author></authors></contributors><auth-address>太原理工大學電氣與動力工程學院;</auth-address><titles><title>基于隨機森林與時空聚類的共享單車站點需求量預測</title><secondary-title>科學技術(shù)與工程</secondary-title></titles><periodical><full-title>科學技術(shù)與工程</full-title></periodical><pages>89-94</pages><volume>18</volume><number>32</number><keywords><keyword>隨機森林</keyword><keyword>分層聚類</keyword><keyword>對數(shù)優(yōu)化</keyword><keyword>需求量預測</keyword></keywords><dates><year>2018</year></dates><isbn>1671-1815</isbn><call-num>11-4688/T</call-num><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[11]，此外還受基于2015年美國灣區(qū)70號站點的自行車需求量數(shù)據(jù)，研究了時間因子、氣象因子以及關(guān)聯(lián)站點對需求量的影響，建立了基于隨機森林與時空聚類的模型，實現(xiàn)了對有樁自行車需求量的預測。Li等ADDINEN.CITE<EndNote><Cite><Author>Li</Author><Year>2015</Year><RecNum>123</RecNum><DisplayText><styleface="superscript">[12]</style></DisplayText><record><rec-number>123</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">123</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>YexinLi</author><author>YuZheng</author><author>HuichuZhang</author><author>LeiChen</author></authors></contributors><titles><title>Trafficpredictioninabike-sharingsystem</title><secondary-title>AdvancesinGeographicInformationSystems</secondary-title></titles><periodical><full-title>AdvancesinGeographicInformationSystems</full-title></periodical><keywords><keyword>Distanceoracle</keyword><keyword>Roadnetwork</keyword><keyword>Spatialanalyticalquery</keyword><keyword>Systemarchitecture</keyword></keywords><dates><year>2015</year></dates><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[12]提出了一種分層預測模型，運用二分聚類算法和漸變增強回歸樹模型來預測站點的借還車數(shù)量。對于預測方法，關(guān)于調(diào)度需求量預測研究所使用的方法主要有傳統(tǒng)時間序列分析方法、機器學習模型及深度學習算法。傳統(tǒng)時間序列分析方法如差分整合移動平均自回歸模型(autoregressiveintegratedmovingaveragemodel，ARIMA)、多元回歸分析、馬爾可夫鏈等模型，是最早被應(yīng)用到共享單車需求預測的一類方法。Kaltenbrunner等ADDINEN.CITE<EndNote><Cite><Author>Kaltenbrunner</Author><Year>2010</Year><RecNum>122</RecNum><DisplayText><styleface="superscript">[13]</style></DisplayText><record><rec-number>122</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">122</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>AndreasKaltenbrunner</author><author>RodrigoMeza</author><author>JensGrivolla</author><author>JoanCodina</author><author>RafaelBanchs</author></authors></contributors><auth-address>BarcelonaMedia-InnovationCentre,Diagonal177,planta9,08018Barcelona,Spain</auth-address><titles><title>Urbancyclesandmobilitypatterns:Exploringandpredictingtrendsinabicycle-basedpublictransportsystem</title><secondary-title>PervasiveandMobileComputing</secondary-title></titles><periodical><full-title>PervasiveandMobileComputing</full-title></periodical><volume>6</volume><number>4</number><keywords><keyword>Mobilitypattern</keyword><keyword>Communitybicycleprogram</keyword><keyword>Urbanbehavior</keyword></keywords><dates><year>2010</year></dates><isbn>1574-1192</isbn><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[13]基于巴塞羅那社區(qū)自行車項目某站點的數(shù)據(jù)，運用ARIMA模型，對可用自行車的數(shù)量進行了預測。閆廈ADDINEN.CITE<EndNote><Cite><Author>閆廈</Author><Year>2018</Year><RecNum>118</RecNum><DisplayText><styleface="superscript">[14]</style></DisplayText><record><rec-number>118</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843880">118</key></foreign-keys><ref-typename="Thesis">32</ref-type><contributors><authors><author>閆廈</author></authors><tertiary-authors><author>尹寶才,</author></tertiary-authors></contributors><titles><title>基于站點聚類的公共自行車系統(tǒng)需求量預測</title></titles><keywords><keyword>公共自行車系統(tǒng)</keyword><keyword>站點聚類算法</keyword><keyword>需求量預測</keyword><keyword>SARIMA模型</keyword><keyword>Lasso回歸</keyword></keywords><dates><year>2018</year></dates><publisher>大連理工大學</publisher><work-type>碩士</work-type><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[14]根據(jù)單車需求量的時序性，建立了考慮季節(jié)周期的ARIMA模型，該模型可以刻畫出行需求的周期性和趨勢性。盡管ARIMA等統(tǒng)計推斷模型在時間序列建模中顯示出一定的有效性，但是無法刻畫需求量與各影響因素之間的時空依賴性等復雜非線性關(guān)系。而且，實際應(yīng)用中數(shù)據(jù)的噪聲會降低參數(shù)估計的可靠性，因而預測效果不是特別理想。近年來，隨著海量出行數(shù)據(jù)的積累和計算能力的提高，利用機器學習方法發(fā)現(xiàn)交通系統(tǒng)的動態(tài)特性逐漸成為一個研究熱點。支持向量回歸(supportvectorregression，SVR)、隨機森林(randomforest，RF)和神經(jīng)網(wǎng)絡(luò)(neuralnetworks，NN)的模型已廣泛用于共享單車的短時需求預測。根據(jù)無樁式共享單車需求量的時間序列特征，孔靜ADDINEN.CITE<EndNote><Cite><Author>孔靜</Author><Year>2018</Year><RecNum>116</RecNum><DisplayText><styleface="superscript">[3]</style></DisplayText><record><rec-number>116</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1621843256">116</key></foreign-keys><ref-typename="Thesis">32</ref-type><contributors><authors><author>孔靜</author></authors><tertiary-authors><author>胡大偉,</author></tertiary-authors></contributors><titles><title>無樁式共享單車站點需求預測及調(diào)度路徑優(yōu)化研究</title></titles><keywords><keyword>無樁式共享單車</keyword><keyword>BP神經(jīng)網(wǎng)絡(luò)</keyword><keyword>需求預測</keyword><keyword>調(diào)度路徑優(yōu)化</keyword><keyword>混合算法</keyword></keywords><dates><year>2018</year></dates><publisher>長安大學</publisher><work-type>碩士</work-type><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[3]建立了基于BP神經(jīng)網(wǎng)絡(luò)的預測方法模型，由于缺乏對天氣、位置等外部影響因素的建模，預測效果并不理想。曹旦旦ADDINEN.CITE<EndNote><Cite><Author>曹旦旦</Author><Year>2021</Year><RecNum>90</RecNum><DisplayText><styleface="superscript">[15]</style></DisplayText><record><rec-number>90</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824463">90</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>曹旦旦</author><author>范書瑞</author><author>夏克文</author></authors></contributors><auth-address>河北工業(yè)大學電子信息工程學院;河北工業(yè)大學大數(shù)據(jù)重點實驗室;</auth-address><titles><title>共享單車短時需求量預測的機器學習方法比較</title><secondary-title>計算機仿真</secondary-title></titles><periodical><full-title>計算機仿真</full-title></periodical><pages>92-97</pages><volume>38</volume><number>01</number><keywords><keyword>共享單車</keyword><keyword>數(shù)據(jù)分析</keyword><keyword>極端隨機樹</keyword><keyword>需求量預測</keyword><keyword>機器學習</keyword></keywords><dates><year>2021</year></dates><isbn>1006-9348</isbn><call-num>11-3724/TP</call-num><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[15]采用隨機森林、極端隨機樹、支持向量機、人工神經(jīng)網(wǎng)絡(luò)、XGBoost這5種機器學習方法，實現(xiàn)對共享單車短時需求量的預測，結(jié)果顯示極端隨機樹預測效果最優(yōu)。華明壯ADDINEN.CITE<EndNote><Cite><Author>華明壯</Author><Year>2018</Year><RecNum>88</RecNum><DisplayText><styleface="superscript">[6]</style></DisplayText><record><rec-number>88</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824448">88</key></foreign-keys><ref-typename="Thesis">32</ref-type><contributors><authors><author>華明壯</author></authors><tertiary-authors><author>陳學武,</author></tertiary-authors></contributors><titles><title>基于訂單數(shù)據(jù)挖掘的共享單車調(diào)度需求分析方法研究</title></titles><keywords><keyword>共享單車</keyword><keyword>訂單數(shù)據(jù)</keyword><keyword>出行特征</keyword><keyword>聚類分析</keyword><keyword>客流預測</keyword><keyword>調(diào)度需求</keyword></keywords><dates><year>2018</year></dates><publisher>東南大學</publisher><work-type>碩士</work-type><urls></urls><remote-database-provider>Cnki</remote-database-provider></record></Cite></EndNote>[6]采用多種方法進行預測，線性方法使用歷史均值和時間序列，非線性方法使用神經(jīng)網(wǎng)絡(luò)和隨機森林。用平均絕對誤差、均方根誤差、R方等指標，評估各種預測方法的優(yōu)劣。結(jié)果發(fā)現(xiàn)，隨機森林預測結(jié)果最好。陳星佑通過聚類LSTM與ARIMA模型、支持向量機（回歸）、未使用簇聚類的LSTM模型的預測效果的比較，驗證了聚類LSTM模型的有效性。徐長興ADDINEN.CITE<EndNote><Cite><Author>徐長興</Author><Year>2021</Year><RecNum>89</RecNum><DisplayText><styleface="superscript">[16]</style></DisplayText><record><rec-number>89</rec-number><foreign-keys><keyapp="EN"db-id="d9t90r2wpfz9dle22z3vw2z2ee9av0r5z22p"timestamp="1620824463">89</key></foreign-keys><ref-typename="JournalArticle">17</ref-type><contributors><authors><author>徐長興</author><author>汪偉平</author><author>昌錫銘</author><author>包旭</author><author>吳建軍</author></authors></contributors><auth-address>北京交通大學軌道交通控制與安全國家重點實驗室;淮陰工學院交通工程學院;</auth-address><titles><title>基于因果分析和相似日選擇的共享單車需求量預測組合模型</title><secondary-title>山東科學</secondary-title></titles><periodical><full-title>山東科學</full-title></periodical><pages>54-64</pages><volume>34</volume><number>02</number><keywords><keyword>共享單車</keyword><keyword>出行需求</keyword><keyword>因果分析</keyword><keyword>灰色關(guān)聯(lián)度</keyword><keyword>相似日</keyword><keyword>機器學習</keyword><keyword>Stacking策略</keyword></keywords><dates><year>2021</year></dates><isbn>1002-4026</isbn><call-num>37-1188/N</call-num><urls></urls><remote-database-provider>Cnki</remote-database-provider></record