ʻO Dmitry Kazakov, Ke alakaʻi pūʻulu ʻikepili ʻikepili ma Kolesa Group, kaʻana like i nā ʻike mai ka loiloi Kazakhstan mua o nā ʻoihana ʻikepili.
Ma ke kiʻi: Dmitry Kazakov
E hoʻomanaʻo i ka ʻōlelo kaulana ʻo Big Data e like me ka wahine ʻōpio - ke kamaʻilio nei nā mea a pau e pili ana iā ia, akā ʻaʻohe mea i ʻike inā aia maoli ia. Hiki ke ʻōlelo ʻia e pili ana i ka mākeke no nā loea ʻikepili (ma Kazakhstan) - aia ka hype, akā ʻo wai ka mea ma hope o ia (a inā aia kekahi ma laila) ʻaʻole maopopo loa - ʻaʻole i HR, ʻaʻole hoʻi i nā luna, ʻaʻole hoʻi. nā ʻepekema ʻikepili iā lākou iho.
Ua lilo mākou , kahi i noiʻi ai lākou ma mua o 300 mau loea e pili ana i kā lākou uku, nā hana, nā mākau, nā mea hana a me nā mea hou aku.
Spoiler: ʻAe, aia maoli lākou, akā ʻaʻole maʻalahi nā mea āpau.
ʻIke maikaʻi. ʻO ka mea mua, ʻoi aku ka nui o nā ʻepekema data ma mua o kā mākou i manaʻo ai. Ua hiki iā mākou ke nīnauele i 300 poʻe, ʻaʻole i waena o lākou wale nō nā huahana, kūʻai aku a me nā mea loiloi BI, akā ʻo ML a me DWH ʻenekini hoʻi, he mea ʻoluʻolu loa. ʻO ka hui nui loa ka poʻe i kapa iā lākou iho he ʻepekema data - ʻo ia ka 36% o nā mea pane. He mea paʻakikī ke haʻi inā pili kēia i ka noi o ka mākeke a ʻaʻole paha, no ka mea ke kūkulu ʻia nei ka mākeke ponoʻī.
Huikau ka mahele ana o na pae hana - ua aneane like ka nui o na alakai a me na luna hui e like me na opiopio. Nui paha nā kumu no kēia. No ka laʻana, he helu nui o nā hui liʻiliʻi o 2-3 poʻe, kahi e hiki ai i ke alakaʻi ke lilo i waena waena a i ʻole ka pae kiʻekiʻe.
ʻO kekahi kumu ʻē aʻe paha ka haunaele e noho aliʻi nei i ka mākeke e pili ana i nā kūlana i ka hāʻawi ʻana i nā kuleana a me nā hana. Hāʻawi ʻia nā alakaʻi hui i kekahi manawa i ka poʻe e hana wale ana i hoʻokahi makahiki a ʻelua paha ʻoi aku ka lōʻihi ma mua o nā poʻe ʻē aʻe, me ka ʻole e pili ana i ke kiʻekiʻe o ka mākaukau a me ka ʻike. ʻIke mākou i kēia ma ka hāʻawi ʻana i nā hana ma ke kūlana - 38% o nā mana a me nā alakaʻi o ka hui e hana nei i ka hana mua a me kekahi 33% i ka loiloi helu kumu.
Maanei ua noi mākou i nā mea pane e loiloi i ka pae o ka analytics i kā lākou ʻoihana. Inā ʻoe e nānā pono, hiki iā ʻoe ke ʻike i ka 10% o ka poʻe pane e hana nei i nā keʻena analytics o 2-3 mau kānaka e manaʻoʻiʻo he "pae kiʻekiʻe" kā lākou.
He aha ka "pae kiʻekiʻe"? Hana maikaʻi ka ʻōnaehana BI. Aia ka DWH a me ka ʻikepili nui. Hana mau ʻia nā hoʻokolohua A/B. Aia nā ʻōnaehana ML a me DS i ka hana. Hoʻoholo ʻia ma muli o ka ʻikepili wale nō. ʻO ka ʻoihana ʻikepili a me ka ʻepekema data kekahi o nā mea nui i ka ʻoihana.
Aneane hiki ʻole ke hoʻokō i nā mea āpau i luna me kahi keʻena o 2-3 poʻe. Ke manaʻo nei au he mea ulu iki kēia hopena o ka noiʻi - ʻaʻohe kanaka e hoʻohālikelike iā lākou iho i mea e hoʻoholo pono ai i ko lākou pae.
E like me ka mea i manaʻo ʻia, hoʻohana ka poʻe ʻepekema data i ka hapa nui o ko lākou manawa ʻaʻole ma ka makemakika paʻakikī a i ʻole ka ʻenekinia, akā ma ka preprocessing, hoʻoiho a me ka hoʻomaʻemaʻe ʻana i ka ʻikepili. Ma kēlā me kēia ʻoihana ʻike mākou i ka preprocessing i ka 3 kiʻekiʻe. Akā ʻaʻole mākou e ʻike i nā mea paʻakikī e like me ka hoʻomohala ʻana i nā hiʻohiʻona ML a i ʻole ka hana ʻana me Big Data ma ka 3 kiʻekiʻe - ma waena wale nō o nā ʻenekinia ML a me DWH.
Aia kekahi mau ʻike kaumaha. Hoʻonohonoho ka poʻe loea i ka 40% o kā lākou mau hana iā lākou iho. Ma Kazakhstan, ʻo nā hui unicorn kiʻekiʻe wale nō i hoʻāʻo i nā pono o ka hana ʻana me ka ʻikepili nui a aʻo pehea e hana pono ai. Hoʻolaha lākou i ka mākeke ʻo Big Data a me Machine Learning he ʻoluʻolu, a ke hahai nei ka lua o ka echelon ma hope, akā ʻaʻole maopopo mau i ka hana ʻana me ka ʻikepili. No laila, ʻike mākou ua hoʻonohonoho nā loea i nā hana no lākou iho, a ʻaʻole ʻike mau nā ʻoihana i kā lākou makemake.
Pīhoihoi au ʻaʻole ʻike ʻo 20% o nā loea inā he Data Warehouse kā lākou hui. ʻAe, a me nā ʻōnaehana hoʻokele waihona ʻaʻole maikaʻi loa nā mea a pau - 41% hoʻohana MySQL, a me kekahi 34% hoʻohana i ka PostgreSQL. He aha ke ʻano o kēia? Hana lākou me nā ʻikepili liʻiliʻi.
I ka nīnau e pili ana i nā ʻōnaehana mālama, ʻike hou mākou i ka MySQL a me (!) Excel. Akā e hōʻike paha kēia, no ka laʻana, ʻaʻole i loaʻa i ka hapa nui o nā ʻoihana kahi noi e hana me ka ʻikepili nui.
Eia hou nā mea a pau. Ma keʻano laulā, ua emi iki nā uku ma mua o kaʻu i manaʻo ai.
ʻO wau iho, paʻakikī iaʻu e noʻonoʻo i kahi ʻenekini ML i mākaukau e hana no 200 tausani tenge - he intern paha ʻo ia. He nāwaliwali loa nā mākau o ia mau loea, a i ʻole paʻakikī i nā ʻoihana e loiloi pono i ka hana o ka ʻIkepili ʻIkepili. Akā paha e hōʻike ana kēia i ka mākeke i ka hoʻomaka ʻana o kona oʻo ʻana. A i ka wā lōʻihi, e hoʻokumu ʻia ka pae o nā uku ma kahi pae kūpono.
Source: www.habr.com