2.4.2 Ukubikezela kanye nowcasting
Ukubikezela ikusasa kunzima, kodwa ukubikezela samanje okulula.
Abacwaningi bamasu ababili abakhulu abangasebenzisa ngedatha yokubuka ibikezela . Ukwenza ukuqagela ngekusasa kubonakala kuyinkimbinkimbi, futhi mhlawumbe ngenxa yalesi sizathu, ukubikezela okwamanje akuyona ingxenye enkulu yocwaningo lomphakathi (nakuba kuyincenye ebalulekile futhi ebalulekile ekudalweni kwemibono yabantu, ezomnotho, epidemiology kanye nesayensi yezombusazwe). Kodwa lapha, ngingathanda ukugxila ekuhloleni okukhethekile okubizwa ngokuthi i- nowcasting -igama elisuka ekuhlanganiseni "manje" nangokuthi "ukubikezela." Esikhundleni sokubikezela ikusasa, ukuzamazama ukuzama ukusebenzisa imibono ngokubikezela ukukala isimo samanje wezwe; lizama "ukubikezela okwamanje" (Choi and Varian 2012) . I-Nowcasting inamandla okuba usizo ikakhulukazi kuhulumeni nezinkampani ezidinga izinyathelo ezifike ngesikhathi futhi ezinembile zomhlaba.
Omunye obeka lapho isidingo sokulinganisa okufika ngesikhathi nesifanele kucace kakhulu yi-epidemiology. Cabanga ngesimo somkhuhlane ("umkhuhlane"). Unyaka ngamunye, izifo zengculaza zonyaka zidala izigidi zezifo namakhulu ezinkulungwane zabantu abafa emhlabeni jikelele. Ngaphezu kwalokho, unyaka ngamunye, kungenzeka ukuthi uhlobo lwe-novel lwangase luvele luyobulala izigidi. Ngokwesibonelo, ukukhukhumeza komkhuhlane we-1918, kulinganiselwa ukuthi kwabulala abantu abaphakathi kuka-50 no-100 million (Morens and Fauci 2007) . Ngenxa yesidingo sokulandelela futhi okungasabela ekuqhumeni komkhuhlane, ohulumeni emhlabeni wonke baye badala izinhlelo zokubheja komkhuhlane. Isibonelo, i-US Centers for Disease Control and Prevention (CDC) njalo futhi iqoqa ulwazi oluvela odokotela abakhethwe ngokucophelela ezweni lonke. Nakuba le nqubo ikhiqiza idatha ephezulu, ine-lag yokubika. Lokhu kungukuthi, ngenxa yesikhathi esithathayo ukuthi idatha ifike kudokotela ukuze ihlanzwe, isetshenziswe futhi ishicilelwe, uhlelo lwe-CDC lukhipha ukulinganiswa kokuthi uketshezi oluningi lwalukhona amasonto amabili edlule. Kodwa, lapho sibhekene nesifo esiwumshayabhuqe, izikhulu zezempilo zomphakathi azifuni ukuthi umkhuhlane ukhona kangakanani amasonto amabili edlule; bafuna ukwazi ukuthi umkhuhlane ukhona kangakanani manje.
Ngesikhathi esifanayo ukuthi i-CDC iqoqa idatha ukulandelela umkhuhlane, i-Google iphinda iqoqe idatha mayelana nesandulela se-influenza, nakuba ifomu ehlukile kakhulu. Abantu abavela emhlabeni wonke bahlala bethumela imibuzo ku-Google, futhi eminye yale mibuzo-njengokuthi "imishanguzo yesifo sofuba" kanye "nezibonakaliso zemfuluwenza" -ngabonisa ukuthi umuntu owenza umbuzo unomkhuhlane. Kodwa, ukusebenzisa le mibuzo yokucwaninga ukulinganisa ukusabalala komkhuhlane kuyinkimbinkimbi: akuwona wonke umuntu onomkhuhlane owenza ukusesha okuhlobene nomkhuhlane, futhi akuwona wonke usesho oluhlobene nokukhuhlane oluvela kumuntu onomkhuhlane.
UJeremy Ginsberg kanye neqembu labalingani bakhe (2009) , abanye ku-Google nakwezinye i-CDC, babenombono obalulekile futhi ohlakaniphile ukuhlanganisa lemithombo emibili yedatha. Cishe, ngokusebenzisa uhlobo lwe-alchemy yezibalo, abacwaningi bahlanganisa idatha yokucinga okusheshayo nengalungile nedatha ye-CDC ehamba kancane futhi enembile ukuze kutholakale isilinganiso esisheshayo nesinembile sokukhula komkhuhlane. Enye indlela yokucabanga ngayo ukuthi basebenzisa idatha yokusesha ukusheshisa idatha ye-CDC.
Ngokuqondile, besebenzisa idatha kusukela ngo-2003 kuya ku-2007, i-Ginsberg kanye nozakwethu bacatshangelwa ukuthi ubuhlobo obuphakathi kokusakazeka komkhuhlane ku-CDC yedatha kanye nenani lokusesha ngemigomo ehlukene yezigidi ezingu-50. Kule nqubo, eyayiqhutshwa ngokugcwele kwedatha futhi ayidingi ulwazi olukhethekile lwezokwelapha, abacwaningi bathola imibuzo ehlukene engu-45 eyabonakala iyinhloko yokudlulela kwedatha yokukhula komkhuhlane we-CDC. Khona-ke, ngokusebenzisa ubudlelwane ababafunde kusukela ku-2003 kuya ku-2007, uGinsberg kanye nosebenza nabo bahlolisisa imodeli yabo phakathi nenkathi ka-2007-2008 yemithonjane. Bathola ukuthi izinqubo zabo zingenza ngempela izicucu eziwusizo futhi ezinembile (isibalo 2.6). Le miphumela yashicilelwe ku- Nature futhi ithola ukufakwa kwezindaba zokukhangisa. Le phrojekthi-ebizwa ngokuthi Amathrendi we-Google Flu-yaba umdwebo ophindaphindiwe ngokuphindaphindiwe ngamandla wedatha enkulu yokushintsha umhlaba.
Umdwebo 2.6: uJeremy Ginsberg kanye nozakwethu (2009) bahlangene idatha ye-Google yokusesha ngeCDC yedatha ukudala Amathrendi we-Google Flu, okwakungakhipha izinga lokugula okunjengomkhuhlane (ILI). Imiphumela kulesi sibalo yilapho ephakathi kwe-Atlantic esifundeni sase-United States ngesikhathi sonyaka we-2007-2008 we-influenza. Nakuba ekuqaleni kwakuthembise kakhulu, ukusebenza kweThrough Trends ye-Google kwalimala ngokuhamba kwesikhathi (Cook et al. 2011; Olson et al. 2013; Lazer et al. 2014) . Ilandelwe kusuka ku- Ginsberg et al. (2009) , isibalo 3.
Kodwa-ke, le ndaba ebonakalayo impumelelo yagcina yaba yilahla. Ngokuhamba kwesikhathi, abacwaningi bathola ukulinganiselwa okubili okubalulekile okwenza amathrendi e-Google Flu abe mnandi kunokuba avele ekuqaleni. Okokuqala, ukusebenza kweThint Trends ye-Google Flu empeleni kwakungcono kangcono kakhulu kuneyomodeli elula esilinganisa inani lomkhuhlane osuselwa ekuqaliseni okungafani okuvela emiphakathini emibili yamuva yokukhulelwa komkhuhlane (Goel et al. 2010) . Futhi, ngezikhathi ezithile, i-Google Flu Trends yayiyibi kakhulu kunalendlela elula (Lazer et al. 2014) . Ngamanye amazwi, Amathrendi we-Google Flu nazo zonke idatha zawo, ukufundwa komshini, kanye nekhompyutha enamandla akukhambelanga ngokuphawulekayo ukulula okulula nokuzwakala kalula. Lokhu kusikisela ukuthi uma uhlola noma yikuphi ukubikezela noma ukusabalalisa, kubalulekile ukuqhathanisa ngokusisekelo.
I-caveat yesibili ebalulekile mayelana nezendlela ze-Google Flu yukuthi amandla ayo okubikezela ukuthi idatha ye-flux ye-CDC yayijwayele ukuhluleka kwesikhashana kanye nokubola kwesikhathi eside ngenxa yokuqhuma nokuguqulwa kwe- algorithmic . Isibonelo, phakathi no-2009 we-Swine Flu ukuqubuka kweziThambeko ze-Google Flu Trends ngokuphawulekayo kuveze inani lemfuluwenza, mhlawumbe ngoba abantu bathambekele ekuguquleni ukuziphatha kwabo kokusesha ngenxa yokusatshalaliswa kwebhubhane emhlabeni jikelele (Cook et al. 2011; Olson et al. 2013) . Ngaphezu kwalezi zinkinga zesikhashana, ukusebenza kwancipha kancane kancane ngokuhamba kwesikhathi. Ukuthola izizathu zalokhu ukubola kwesikhathi eside kunzima ngoba ukulungiswa kwe-Google search izici, kodwa kubonakala sengathi ngo-2011 i-Google yaqala ukuphakamisa amagama afanayo osesho lapho abantu befuna izimpawu zofuzo ezifana "nomkhuhlane" nokuthi "ukukhwehlela" (kubonakala sengathi lesi sici asisasebenzi). Ukwengeza lesi sici kuyinto enengqondo ukuyenza uma usebenzisa injini yokusesha, kodwa lokhu kuguqulwa kwe-algorithmic kunomthelela wokwenza usesho oluthile lwezempilo olwenza ukuthi Amathrendi we-Google avele ekukhuliseni ukukhulelwa komkhuhlane (Lazer et al. 2014) .
Lezi zigcawu ezimbili ziphoqelela imizamo yekusasa ezayo, kodwa ayibhubhisi. Empeleni, ngokusebenzisa izindlela eziqaphile, Lazer et al. (2014) Yang, Santillana, and Kou (2015) bakwazi ukugwema lezi zinkinga ezimbili. Ukuqhubekela phambili, ngilindele ukuthi izifundo ezikhona manje ezihlanganisa imithombo emikhulu yedatha kanye nedatha eqoqiwe nomcwaningi izokwenza izinkampani kanye nohulumeni bakwazi ukudala izilinganiso ezihamba phambili nangesikhathi esinembile ngokusheshisa noma yikuphi ukulinganisa okwenziwe ngokuphindaphindiwe ngokuhamba kwesikhashana. Amaphrojekthi we-Nowcasting afana namathrendi e-Google Flu akhombisa ukuthi yini engenzeka uma imithombo emikhulu yedatha ihlanganiswa nemininingwane eminingi yendabuko eyenziwe ngezinhloso zocwaningo. Uma ucabanga ngokufana nesithombe sokuqala kwesahluko 1, ukusakazwa kwamanje kunamandla okuhlanganisa ama-readymades wesitayela se-Duchamp nama-post-style wesigcawu se-Michelangelo ukuze ahlinzeke abenzi izinqumo ngezilinganiso ezengeziwe ngesikhathi esifanele futhi ezinembile zesikhathi samanje nezibikezelo esikhathini esizayo esiseduze.
