Dongosolo Lopangidwa ndi AI, losavuta komanso lofulumira kupeza deta yapamwamba kwambiri?

AI idapanga data yopangira muzochita

Syntho, katswiri wazopanga zopangidwa ndi AI, akufuna kutembenuka privacy by design kukhala mwayi wampikisano ndi data yopangidwa ndi AI. Amathandizira mabungwe kupanga maziko olimba a data ndi mwayi wosavuta komanso wachangu ku data yapamwamba kwambiri ndipo posachedwapa adapambana Mphotho ya Philips Innovation.

Komabe, kupanga zidziwitso zopangidwa ndi AI ndi njira yatsopano yomwe imayambitsa mafunso omwe amafunsidwa pafupipafupi. Kuti tiyankhe izi, Syntho adayambitsa phunziro limodzi ndi SAS, mtsogoleri wamsika mu Advanced Analytics ndi pulogalamu ya AI.

Pogwirizana ndi Dutch AI Coalition (NL AIC), adafufuza zamtengo wapatali wa deta yopangidwa poyerekezera deta yopangidwa ndi AI yopangidwa ndi Syntho Engine ndi deta yoyambirira kudzera muzofufuza zosiyanasiyana pa khalidwe la deta, kuvomerezeka kwalamulo ndi kugwiritsidwa ntchito.

Kodi kusadziwika kwa data si njira yothetsera?

Njira zachikale zosadziwika bwino ndizofanana zomwe zimasokoneza zomwe zidachitika kale kuti zilepheretse kutsata anthu. Zitsanzo ndi kuphatikiza, kupondereza, kufufuta, kutchula mayina, kubisa deta, komanso kusanja mizere ndi mizere. Mukhoza kupeza zitsanzo mu tebulo ili m'munsimu.

data anonymization

Njirazi zimabweretsa zovuta zitatu zazikulu:

  1. Amagwira ntchito mosiyana pamtundu uliwonse wa data komanso pamtundu uliwonse, kuwapangitsa kukhala ovuta kukulitsa. Komanso, popeza amagwira ntchito mosiyana, nthawi zonse pamakhala mkangano wokhudza njira zomwe zingagwiritsidwe ntchito komanso kuphatikiza kwa njira zomwe zikufunika.
  2. Nthawi zonse pamakhala ubale umodzi ndi umodzi ndi deta yoyambirira. Izi zikutanthauza kuti nthawi zonse padzakhala chiwopsezo chachinsinsi, makamaka chifukwa cha ma dataset otseguka ndi njira zomwe zilipo zolumikizira ma datasetwo.
  3. Iwo amasokoneza deta ndipo potero kuwononga deta mu ndondomeko. Izi ndizowononga kwambiri ntchito za AI kumene "mphamvu yolosera" ndiyofunikira, chifukwa deta yoyipa idzabweretsa malingaliro oipa kuchokera ku chitsanzo cha AI (Zinyalala-mu zidzabweretsa zinyalala).

Mfundozi zikuwunikidwanso kudzera mu phunziro ili.

Chiyambi cha phunzirolo

Pakafukufuku wamilanduyo, deta yomwe mukufuna kutsata inali matelefoni operekedwa ndi SAS omwe ali ndi makasitomala 56.600. Setiyi ili ndi magawo 128, kuphatikiza ndime imodzi yosonyeza ngati kasitomala wachoka pakampani (ie 'churned') kapena ayi. Cholinga cha phunziroli chinali kugwiritsa ntchito deta yopangira kuti aphunzitse zitsanzo zina kuti ziwonetsere kusintha kwa makasitomala ndikuwunika momwe machitidwe ophunzitsidwawo akuyendera. Monga kulosera kwa churn ndi ntchito yamagulu, SAS idasankha mitundu inayi yodziwika bwino kuti ineneretu, kuphatikiza:

  1. Nkhalango zokhazikika
  2. Kuwonjezeka kwa gradient
  3. Kutsika kwa mayendedwe
  4. Neural network

Asanapange deta yopangira, SAS inagawaniza dataseti ya telecom mwachisawawa kukhala sitima yapamtunda (pophunzitsa ma model) ndi seti yotsekera (popeza zitsanzo). Kukhala ndi malo osungiramo zigoli kumathandizira kuwunika kosakondera komwe mtundu wamagulu ungagwire bwino ukagwiritsidwa ntchito pazatsopano.

Pogwiritsa ntchito sitimayi monga zolowetsa, Syntho adagwiritsa ntchito Syntho Engine yake kupanga deta yopangira. Pakuyika chizindikiro, SAS idapanganso masitima apamtunda osinthidwa atagwiritsa ntchito njira zingapo zozindikiritsa anthu kuti afikire malire ena (a k-anonimity). Masitepe akale adapanga ma dataset anayi:

  1. Seti ya data ya sitima yapamtunda (ie dataset yoyambirira kuchotsera dataset)
  2. Seti ya data yosungidwa (ie kagawo kakang'ono ka dataset yoyambirira)
  3. Seti ya data yosadziwika (kutengera data ya sitima yapamtunda)
  4. Seti ya data yopangira (yotengera mayendedwe a sitima yapamtunda)

Ma dataset 1, 3 ndi 4 adagwiritsidwa ntchito pophunzitsa mtundu uliwonse wamagulu, zomwe zidapangitsa kuti pakhale ma 12 (3 x 4) ophunzitsidwa bwino. Pambuyo pake, SAS idagwiritsa ntchito dataset kuti ayeze kulondola komwe mtundu uliwonse umaneneratu kusintha kwamakasitomala. Zotsatira zaperekedwa pansipa, kuyambira ndi ziwerengero zoyambira.

Makina ophunzirira makina opangidwa ku SAS

Chithunzi: Mapaipi ophunzirira pamakina opangidwa mu SAS Visual Data Mining ndi Machine Learning

Ziwerengero zoyambira pofananiza data yosadziwika ndi yoyambirira

Njira zosadziwika bwino zimawononga ngakhale machitidwe oyambira, malingaliro abizinesi, maubale ndi ziwerengero (monga momwe zilili pansipa). Kugwiritsa ntchito deta yosadziwika kwa ma analytics oyambira kumabweretsa zotsatira zosadalirika. M'malo mwake, kusakwanira kwa data yosadziwika kudapangitsa kuti zikhale zosatheka kuzigwiritsa ntchito pazinthu zapamwamba zowunikira (monga AI/ML modelling ndi dashboarding).

kuyerekeza deta yosadziwika ndi yoyambirira

Ziwerengero zoyambira poyerekeza zopangira ndi zoyambira

Kupanga zidziwitso zopanga ndi AI kumasunga machitidwe oyambira, malingaliro abizinesi, maubale ndi ziwerengero (monga momwe zilili pansipa). Kugwiritsa ntchito deta yopangira ma analytics oyambira motero kumatulutsa zotsatira zodalirika. Funso lofunikira, kodi data yopangidwa imakhala ndi ntchito zapamwamba zowunikira (monga AI/ML modelling ndi dashboarding)?

kuphatikiza deta yopangira ku data yoyamba

Zambiri zopangidwa ndi AI komanso kusanthula kwapamwamba

Zopangapanga sizimangokhala pamayendedwe oyambira (monga momwe zasonyezedwera m'magawo akale), zimatengeranso 'zobisika' zakuya zomwe zimafunikira pa ntchito zowunikira zapamwamba. Chotsatirachi chikuwonetsedwa mu bar chart pansipa, kusonyeza kuti kulondola kwa zitsanzo zophunzitsidwa pa data yopangidwa ndi zitsanzo zophunzitsidwa pa deta yoyambirira ndizofanana. Kuphatikiza apo, ndi malo opindika (AUC*) pafupi ndi 0.5, mitundu yophunzitsidwa pa data yosadziwika imachita moyipa kwambiri. Lipoti lathunthu ndi zowunika zonse zapamwamba za analytics pazopanga zopanga poyerekeza ndi zomwe zidayambika zimapezeka popempha.

*AUC: dera lomwe lili pansi pa phirilo ndi muyeso wa kulondola kwa zitsanzo zapamwamba za analytics, poganizira zabwino zenizeni, zabodza, zolakwika zabodza ndi zolakwika zenizeni. 0,5 imatanthawuza kuti zitsanzo zimalosera mwachisawawa ndipo zilibe mphamvu zolosera ndipo 1 imatanthauza kuti chitsanzocho chimakhala cholondola nthawi zonse ndipo chimakhala ndi mphamvu zolosera.

Kuonjezera apo, deta yopangidwirayi ingagwiritsidwe ntchito kumvetsetsa zizindikiro za deta ndi zosintha zazikulu zomwe zimafunikira pa maphunziro enieni a zitsanzo. Zolowetsa zosankhidwa ndi ma algorithms pa data yopangira poyerekeza ndi zomwe zidayambira zinali zofanana kwambiri. Chifukwa chake, njira yowonetsera ikhoza kuchitidwa pamtundu wopangirawu, womwe umachepetsa chiopsezo cha kuphwanya kwa data. Komabe, poyang'ana zolemba za munthu aliyense (monga kasitomala wa telco) kubwerezanso pa deta yoyambirira kumalimbikitsidwa kuti mufotokoze, kuvomereza kowonjezereka kapena chifukwa cha malamulo.                              

AUC ndi algorithm yopangidwa ndi Method

AUC

Zotsatira:

  • Ma Models ophunzitsidwa pa data yopangidwa poyerekeza ndi zitsanzo zophunzitsidwa pa data yoyambirira amawonetsa magwiridwe antchito ofanana kwambiri
  • Ma Models ophunzitsidwa pa data yosadziwika ndi 'njira zachikale zosadziwika bwino' amawonetsa magwiridwe antchito ochepera poyerekeza ndi ma data ophunzitsidwa pa data yoyambirira kapena data yopangira
  • Kupanga ma data opangidwa ndikosavuta komanso mwachangu chifukwa njirayo imagwira ntchito chimodzimodzi pa dataset ndi mtundu wa data.

Njira zogwiritsira ntchito deta yowonjezeretsa mtengo

Gwiritsani ntchito vuto 1: Zambiri zamapangidwe achitsanzo ndi ma analytics apamwamba

Kukhala ndi maziko olimba a data okhala ndi mwayi wosavuta komanso wachangu wogwiritsa ntchito, data yapamwamba kwambiri ndiyofunikira kuti tipange zitsanzo (monga ma dashboards [BI] ndi ma analytics apamwamba [AI & ML]). Komabe, mabungwe ambiri amavutika ndi maziko ocheperako omwe amabweretsa zovuta zazikulu zitatu:

  • Kupeza data kumatenga zaka zambiri chifukwa cha (zachinsinsi) malamulo, njira zamkati kapena ma silo a data
  • Njira zachikale zosadziwika bwino zimawononga deta, zomwe zimapangitsa kuti deta ikhale yosayenerera kusanthula ndi kusanthula kwapamwamba (zinyalala mu = zinyalala kunja)
  • Mayankho omwe alipo si owopsa chifukwa amagwira ntchito mosiyana pamtundu uliwonse wa data komanso mtundu wa data ndipo sangathe kuthana ndi nkhokwe zazikulu zamitundu ingapo.

Njira yopangira data: pangani zitsanzo zokhala ndi data yabwino ngati yeniyeni kuti:

  • Chepetsani kugwiritsa ntchito deta yoyambirira, popanda kulepheretsa opanga anu
  • Tsegulani zidziwitso zanu ndikukhala ndi mwayi wazambiri zomwe zinali zoletsedwa kale (mwachitsanzo chifukwa chachinsinsi)
  • Kufikira kwa data kosavuta komanso kwachangu ku data yoyenera
  • Yankho losavuta lomwe limagwira chimodzimodzi pamasamba aliwonse, nkhokwe zamasamba komanso pamasamba akulu

Izi zimathandiza bungwe kuti lipange maziko olimba a data omwe ali ndi mwayi wosavuta komanso wachangu wogwiritsa ntchito, wapamwamba kwambiri kuti atsegule deta komanso kuti agwiritse ntchito mwayi wa data.

 

Gwiritsani ntchito Mlandu 2: data yoyeserera yanzeru pakuyesa mapulogalamu, kukonza ndi kutumiza

Kuyesa ndi chitukuko ndi deta yoyesera yapamwamba n'kofunika kuti tipereke mayankho amakono a mapulogalamu. Kugwiritsa ntchito deta yoyambirira kumawoneka koonekeratu, koma sikuloledwa chifukwa cha malamulo (zachinsinsi). Njira ina Test Data Management (TDM) zida zoyambitsa "legacy-by-design” pokonza zoyeserera bwino:

  • Osawonetsa zomwe zapanga komanso malingaliro abizinesi ndi kukhulupirika sikusungidwa
  • Gwirani ntchito pang'onopang'ono komanso nthawi yambiri
  • Ntchito yamanja ndiyofunikira

Njira yopangira data: Yesani ndikukhazikitsa ndi data yoyeserera yopangidwa ndi AI kuti mupereke mayankho apulogalamu apamwamba kwambiri ndi:

  • Deta ngati yopanga yokhala ndi malingaliro osungidwa abizinesi ndi kukhulupirika kwawo
  • Kupanga kosavuta komanso kofulumira ndi ma AI apamwamba
  • Zazinsinsi-ndi-mapangidwe
  • Zosavuta, zachangu komanso agile

Izi zimathandiza kuti bungwe liziyesa ndikukula ndi deta yoyeserera yotsatila kuti ipereke mayankho apulogalamu apamwamba!

Zambiri

Wokonda? Kuti mumve zambiri za data yopangidwa, pitani patsamba la Syntho kapena funsani Wim Kees Janssen. Kuti mudziwe zambiri za SAS, pitani www.sas.com kapena lemberani kees@syntho.ai.

Pakugwiritsa ntchito izi, Syntho, SAS ndi NL AIC amagwirira ntchito limodzi kuti akwaniritse zomwe akufuna. Syntho ndi katswiri wazopangidwa ndi AI ndipo SAS ndi mtsogoleri wamsika pazambiri ndipo amapereka mapulogalamu owunikira, kusanthula ndikuwona deta.

* Imaneneratu za 2021 - Njira za Data ndi Analytics kuti Zilamulire, Kukulitsa ndi Kusintha Bizinesi Yapa digito, Gartner, 2020.

syntho guide chivundikiro

Sungani kalozera wanu wazinthu zopangira tsopano!

  • Kodi data yopanga ndi chiyani?
  • Chifukwa chiyani mabungwe amachigwiritsa ntchito?
  • Mtengo wowonjezera wamakasitomala opangira data
  • Momwe mungayambire
Pemphani kalozera
Lumikizanani ndi akatswiri