2.4.2 Forecasting kunye nowcasting
Ukuqikelela elizayo nzima, kodwa ukuqikelela langoku kulula.
Abaphandi beesicwangciso-sibini esibalulekileyo abanokuyisebenzisa ngeenkcukacha zokuqwalasela zibikezela . Ukwenza ukuqwalasela ngekamva kunzima kakhulu, kwaye mhlawumbi ngenxa yoko, ukuxela kwangoku akuyiyona nxalenye enkulu yophando loluntu (nangona luyingxenye encinci kunye ebalulekileyo yokudemokhrasi, i-economics, epidemiology kunye nesayensi yezopolitiko). Nakhu, ke, ndingathanda ukugxininisa uhlobo oluthile lokubikezela olubizwa ngokuba yi- nowcasting -gama elivela ekudibaniseni "ngoku" kunye "nokubikezela." Kunokuba uqikelele ikusasa, ukuzama ukuhlaziya ukuzama ukusebenzisa iinjongo ukusuka ekuqikeleleni ukulinganisa imeko yangoku zehlabathi; uzama "ukuxela kwangoku" (Choi and Varian 2012) . I-Nowcasting inokubaluleka ngakumbi kuburhulumente nakwiinkampani ezifuna amanyathelo afikelelekileyo nakwihlabathi.
Omnye ubeka apho imfuneko yokulinganisa ngexesha elichanekileyo nechanekileyo kucacile ukuba yi-epidemiology. Cinga ngolu hlobo lomkhuhlane ("umkhuhlane"). Unyaka ngamnye, izifo zesifo sengculazi zonyaka zenza izigidi zezifo kunye namakhulu ezinkulungwane zokufa emhlabeni wonke. Ukongezelela, unyaka ngamnye, kunokwenzeka ukuba inokuveliswa kwimoya yemveli yomshushu eya kubulala izigidi. Ukuqhambuka kwe-influenza ye-1918, umzekelo, kuqikelelwa ukuba ubulele phakathi kwabangama-50 ukuya kwi-100 yezigidi zabantu (Morens and Fauci 2007) . Ngenxa yesidingo sokulandelela kunye nokusabela kwisifo sokukhukhumeza umkhuhlane, oorhulumente behlabathi lonke baye badala iinkqubo zokujonga i-influenza. Ngokomzekelo, amaziko ase-United States okuLawula nokuLawula (i-CDC) rhoqo kwaye aqokelele ngolwazi ngokuchanekileyo oogqirha abakhethiweyo belizwe lonke. Nangona le nkqubo ivelisa idatha ephezulu, ine-lag yokubika. Oko kukuthi, ngenxa yexesha elithathayo ukuba idatha ivela koogqirha ukuba ihlambuluke, iqhutywe kwaye ipapashwe, inkqubo ye-CDC ikhupha ukuqikelelwa kobuninzi beentsholongwane kwakukho iiveki ezimbini ezidlulileyo. Kodwa, xa sisingatha isifo esihlakulayo, izikhulu zempilo karhulumente azifuni ukwazi ukuba ininzi imfucuza kwakuphi iiveki ezimbini ezedlulileyo; bafuna ukwazi ukuba ininzi imfucuza ikhoyo ngoku.
Ngexesha elifanayo ukuba i-CDC iqokelele idatha ukuze ilandele umkhuhlane, i-Google iqokelela idatha malunga nokuxhaphaza komkhuhlane, nangona ifom ehlukileyo. Abantu abavela kwihlabathi lonke bahlala bethumela imibuzo kwiGoogle, kwaye ezinye ze mibuzo-ezifana "nezifo zogula" kunye "neempawu zomkhuhlane" -okubonisa ukuba umntu owenza umbuzo unomkhuhlane. Kodwa, ukusebenzisa le mibuzo yokucinga ukuba uqikelele ukuxhaphaza komkhuhlane kukukhohlisayo: akusiyo wonke umntu onomkhuhlane owenza uphando olunefuthe, kwaye akukho uphando olunefuthe luvela kumntu onomkhuhlane.
UJeremy Ginsberg kunye neqela labalingane (2009) , abanye kwiGoogle kunye nabanye kwiCDC, babenombono obalulekileyo kunye neengqondi ukudibanisa le mibini yedatha. Ngokukhawuleza, ngohlobo lwama-alchemy alathisi, abaphandi badibanisa idatha yokukhawuleza nokuchanekileyo kunye nedatha echanekileyo nechanekileyo yeCDC ukwenzela ukuvelisa ngokukhawuleza nokuchanekileyo kwemilinganiselo ye-influenza. Enye indlela yokucinga ngayo kukuba basebenzise idatha yokukhangela ukukhawulezisa idatha yeCDC.
Ngokuchanekileyo, kusetyenziswa idatha ukusuka ngo-2003 ukuya ku-2007, iGinsberg kunye noogxa baqikelelana ubudlelwane obuphakathi kobuninzi be-influenza kwi-CDC yedatha kunye nenani lokutshintshela izigidi ezingama-50. Kusuka kule nkqubo, eyayiqhutywe ngokuchanekileyo kwedatha kwaye ayifuni ulwazi oluthile lwezokwelapha, abaphandi bafumana i-45 yemibuzo eyahlukeneyo eyabonakala iyona nto ibonakalayo kwi-CDC ye-density of data. Emva koko, usebenzisa ubudlelwane abaye bafunda kwi-data 2003-2007, iGinsberg kunye noogxa bavavanya umzekelo wabo ngexesha le-influenza 2007-2008. Bafumanise ukuba iinkqubo zabo zinokubangela ukuba kubekho iifayile ezichanekileyo nezichanekileyo (umfanekiso 2.6). Ezi ziphumo zishicilelwe kwiNdalo kwaye zifunyenwe ngokusasazwa kweendaba. Le projekthi-eyayibizwa ngokuba yi-Google Flu Trends-yaba ngumzekeliso ophindaphindiweyo ngokumalunga nedatha enkulu yokutshintsha ihlabathi.
Umzobo 2.6: UJeremy Ginsberg kunye noogxa (2009) badibanisa idatha ye-Google yophando ngeCDC yedatha ukudala i-Google Flu Trends, engasenza isantya sokugula kwesifo sengculaza (ILI). Iziphumo kulo mfanekiso ziphakathi kommandla wase-Atlantic we-United States kwixesha le-influenza le-2007-2008. Nangona ekuqaleni kwakuthembise kakhulu, ukusebenza kweTGood Flu Trends kwaphazamiseka ngaphezu kwexesha (Cook et al. 2011; Olson et al. 2013; Lazer et al. 2014) . Ukutshintshwa kwi- Ginsberg et al. (2009) , isiqendu 3.
Nangona kunjalo, eli bali eliyimpumelelo elibonakalayo lagqitywa liba yintloni. Ngokuhamba kwexesha, abaphandi bafumene imingcele emibini ebalulekileyo eyenza iGoogle Flu Trends ibe mnandi kunokuba ibonakale kuqala. Okokuqala, ukuqhutyelwa kweendlela zeGoogle Flu okwenene kwakungeyona nto ibhetele kunokuba imodeli elula eqikelela inani lemfuyo esekelwe kwi-extrapolation linear ukususela kwimilinganiselo emibini yentsholongwane yesifo somkhuhlane (Goel et al. 2010) . Kwaye, ngexesha elithile, i-Google Flu Trends yayinzima ngakumbi kunendlela elula (Lazer et al. 2014) . Ngamanye amazwi, i-Google Flu Trends nayo yonke idatha yayo, ukufundwa komatshini kunye nekhomputha enamandla ayizange iphumelele ngokugqithiseleyo ngokulula nokulula ukuyiqonda. Oku kuphakamisa ukuba xa uvavanya nawuphi na umgaqo okanye uhlaselo, kubalulekile ukuthelekisa nxamnye nesiseko.
I-second caveat ebalulekileyo malunga neTGoogle Flu Trends kukuba amandla ayo okuchaza kwangaphambili idatha ye-CDC yomkhuhlane yayilungele ukungaphumeleli kwexesha elifutshane kunye nokubola kwexesha elide ngenxa yokuqhuma kunye ne- algorithmic . Ngokomzekelo, ngexesha le-2009 i-Swine Flu ukubhuka kwe-Google Flu Trends ngokugqithiseleyo kwandisekisa inani lomkhuhlane, mhlawumbi ngenxa yokuba abantu bathambekele ekutshintsheni ukuziphatha kwabo ekutshweni kokusasazeka kwindlala yomhlaba wonke (Cook et al. 2011; Olson et al. 2013) . Ukongeza kwezi ngxaki zexesha elifutshane, ukusebenza kwancipha ngokukhawuleza kwexesha. Ukufumanisa izizathu zolu hlobo lokubola kwexesha elide kunzima kuba i-Google search algorithms i-proprietary, kodwa kubonakala ngathi ngo-2011 i-Google yaqala ukuphakamisa amagama anxulumene nophando xa abantu befuna iimpawu zentsholongwane ezinjenge-"fever" kunye "nokukhwehlela" (kubonakala ngathi le nto ayisasebenza). Ukongeza le nto yinto enengqiqo ukuyenza xa usebenzisa i-injini yokukhangela, kodwa olu tshintsho lwe-algorithmic lunefuthe lokwenza uphando olunxulumene nempilo olubangele ukuqhubela phambili kwe-Google Flu ukuguqulwa kwesifo somkhuhlane (Lazer et al. 2014) .
Ezi zimbini zineengxaki zixakeka kwimigudu yexesha elizayo, kodwa aziyi kutshabalalisa. Enyanisweni, ngokusebenzisa iindlela ezininzi Lazer et al. (2014) , Lazer et al. (2014) kunye Yang, Santillana, and Kou (2015) bakwazi ukuphepha ezi zimbini iingxaki. Ukuqhubela phambili, ndikulindele ukuba izifundo zokufunda ngokutsha zidibanisa imithombo yolwazi ephezulu kunye nedatha efunyenwe ngumphandi iza kwenza iinkampani kunye noorhulumente bakwazi ukudala ukuqikelela okuthe xaxa nangokuchanekileyo ngokukhawulezisa naluphi na umlinganiselo owenziwe ngokuphindaphindiweyo ngexesha elininzi. Iiprojekthi ze-Nowcasting ezifana neTGoogle Flu Trends zibonisa nokuba kwenzeka ntoni ukuba imithombo yolwazi emikhulu idibaniswe nedatha yemveli eyadalwa ngenjongo yophando. Ukucinga emva kobugcisa bezahluko zesahluko soku-1, ukuhlaselwa ngokusasaza kunakho ukuhlanganisa i-Duchamp-style readymades kunye ne-Michelangelo-style custommades ukuze kulungiselelwe ababenzigqibo ngeziphumo ezifanelekileyo ngexesha elizayo kwaye lichanile ngokuchanekileyo kwangoku kunye nokubikezelwa kwexesha elizayo.
