2.3.1.1 Big
Datasets Large ndi yotithandiza; iwo si mapeto mu okha.
The loyamba la makhalidwe atatu deta waukulu kwambiri tinakambirana awa deta lalikulu. Izi magwero deta akhoza kukhala wamkulu mwa njira zitatu zosiyana: anthu ambiri, zambiri mudziwe pa munthu, kapena kuzipenya ambiri pa nthawi. Kukhala ndi gulu lazidziwitso lalikulu limathandiza mitundu yeniyeni heterogeneity kafukufuku kuyeza, kuphunzira zinthu zina kupeza kusiyana yaing'ono, ndi kupanga ziwerengero causal pazimene taona. Izo ukulozeranso kuyambitsa mtundu enieni a sloppiness.
Chinthu choyamba chimene kukula makamaka lipindulitsa ikuyenda yoposa ndi zaka kuti anene kwa subgroups enieni. Mwachitsanzo, Gary King, Jennifer Pan, ndi Molly Roberts (2013) anayeza n'kutheka kuti posts chikhalidwe TV ku China akanati kupimidwa ndi boma. Palokha Mwina izi pafupifupi kufufutidwa si zothandiza kwambiri chifukwa chimene boma censors ena posts koma ena. Koma, chifukwa gulu lazidziwitso awo m'gulu posts 11 miliyoni, Mfumu ndi anzake kubala ziwerengero kwa Mwina kufufuza kwa posts pa magulu 85 osiyana (mwachitsanzo, zolaula, Tibet, ndi magalimoto ku Beijing). Poyerekeza Mwina kufufuza kwa posts m'magulu osiyana, iwo anazindikira zambiri za mmene ndi chifukwa boma censors mitundu ina ya posts. Ndi posts 11 thousand (posts m'malo 11 miliyoni), iwo sakanakhala kubala ziwerengero izi gulu pokha.
Chachiwiri, kukula makamaka lipindulitsa pa akuphunzira wa zinthu kawirikawiri. Mwachitsanzo, Goel ndi anzake (2015) ankafuna kuphunzira njira zosiyanasiyana kuti Tweets mukhoza kupita tizilombo. Chifukwa cascades yaikulu ya kukonzanso Tweets ali kwambiri kawirikawiri ndi pafupifupi mu 3,000 anafunika kuphunzira Tweets oposa biliyoni kuti tipeze cascades opondapo zothandiza.
Chachitatu, datasets lalikulu athe ofufuza kudziwa kusiyana yaing'ono. Ndipotu zambiri za cholinga pa kafukufuku wamkulu mu makampani ndi za kusiyana yaing'ono: molondola kupeza kusiyana mitengo 1% ndi 1.1% pitani kudutsa pa malonda sungathe lakuti mamilioni a madola mu ndalama zoonjezera. Zoikamo ena sayansi, monga kusiyana yaing'ono mwina sangakhale makamaka chofunika (ngakhale powerengera kwambiri). Koma, mu zoikamo ena lamulo, chotero kusiyana ang'onoang'ono angathe kukhala zofunika pamene umaonera akaphatikiza. Mwachitsanzo, ngati pali awiri njira zaumoyo ndi mmodzi pang'ono kwambiri kuposa ena, ndiye kusintha kwa alowererepo aluso akanakhoza kukathera yopulumutsa miyoyo ya anthu ambiri zina.
Pomaliza, lalikulu deta waika kusauka luso lathu kuti anene causal pazimene taona. Ngakhale datasets lalikulu musati wachibadidwe kusintha mavuto ndi kupanga causal kumasulira kwa deta taona, zofanana ndi zachilengedwe zatsopano ziwiri njira zimene ofufuza apanga kupanga zonena causal ku taona data-onse angapindule ndi datasets lalikulu. Ine ndilongosola ndi Mwachitsanzo n'zoona mwatsatanetsatane kenako mu chaputala ichi pamene ine Kufotokoza kafukufuku.
Ngakhale bigness zambiri chuma bwino ntchito molondola, Ndaona kuti bigness zambiri kumabweretsa tchimo mwacilamulo. Pazifukwa zina, bigness Zikuoneka kuti atsogolere ofufuza kunyalanyaza zimene deta yawo kwaiye. Pamene bigness amachita kuchepetsa kufunika nkhawa zolakwa mwachisawawa, kwenikweni kumawonjezera kufunika nkhawa zolakwa mwadongosolo, mitundu zolakwika kuti ine anafotokoza zambiri pansipa kuti chichokera biases mmene deta analengedwa ndi kusonkhanitsidwa. Mu gulu lazidziwitso yaing'ono, onse mwachisawawa zolakwa ndi zolakwa mwadongosolo n'zofunika, koma lalikulu gulu lazidziwitso zolakwa mwachisawawa ali akhoza pakuchitika kutali ndi zolakwa mwadongosolo lalikulu. Akatswiri amene musaganizire zolakwa mwadongosolo adzakhala ntchito datasets yawo yaikulu kupeza kafukufuku ndendende chinthu cholakwika; adzakhala ndendende olakwika (McFarland and McFarland 2015) .
