Dmitry Kazakov, Data Analytics Team Lead paKolesa Group, inogovera ruzivo kubva kuongororo yekutanga yeKazakhstan yevashandi vedata.
Mumufananidzo: Dmitry Kazakov
Rangarira mutsara unozivikanwa wekuti Big Data yakanyanya kufanana nebonde revechiri kuyaruka - munhu wese anotaura nezvazvo, asi hapana anoziva kana iripo. Zvakafanana zvinogona kutaurwa nezve musika wenyanzvi dze data (muKazakhstan) - kune hype, asi ndiani ari shure kwayo (uye kana paine chero munhu aripo zvachose) yakanga isina kunyatsojeka - kana kuna HR, kana maneja, kana data masayendisiti pachavo.
Takapedza , mavakaongorora vanopfuura mazana matatu nyanzvi nezvemihoro yavo, mabasa, hunyanzvi, maturusi nezvimwe zvakawanda.
Spoiler: Hongu, ivo variko zvechokwadi, asi zvese hazvisi nyore.
Kunzwisisa kwakanaka. Chekutanga, kune akawanda data masayendisiti kupfuura zvataitarisira. Takakwanisa kubvunzurudza vanhu 300, pakati pavo vakanga vasiri chete chigadzirwa, kushambadzira uye BI vaongorori, asiwo ML uye DWH mainjiniya, iyo yainyanya kufadza. Boka rakakura kwazvo raisanganisira vese vanozviti masayendisiti data - ndiyo 36% yevakapindura. Zvakaoma kutaura kuti izvi zvinovhara kudiwa kwemusika here kana kuti kwete, nekuti musika pachawo uri kungoumbwa.
Kugovewa kwemazinga ebasa kuri kuvhiringa - kune anenge akawanda etimu anotungamira uye mamaneja sevadiki. Panogona kuva nezvikonzero zvakawanda zveizvi. Semuenzaniso, nhamba huru yezvikwata zviduku zvevanhu 2-3, umo mutungamiri anogona kuva nyanzvi yepakati kana yepamusoro.
Chimwe chikonzero chinogona kunge chiri mhirizhonga iri kutonga pamusika maererano nezviyero mukugoverwa kwemabasa uye kushanda. Vatungamiri vechikwata dzimwe nguva vanopihwa kune avo vanongoshanda kwegore kana maviri kureba kupfuura vamwe, pasina kutaurwa kune nhanho yehunyanzvi uye ruzivo. Izvi tinozviona mukugovaniswa kwemabasa nechinzvimbo - 38% yemaneja nevatungamiriri vechikwata vari kuita pre-processing uye imwe 33% mubasic statistical analysis.
Pano takabvunza vakapindura kuti vatarise nhanho yeanalytics mumakambani avo. Kana ukanyatsotarisisa, unogona kuona kuti 10% yevakapindura vanoshanda mu analytics madhipatimendi e2-3 vanhu vanotenda kuti vane "yepamusoro nhanho."
Chii chinonzi "advanced level"? Iyo BI system inoshanda zvakanaka. Kune DWH uye Big Data. A/B bvunzo dzinoitwa nguva nenguva. Kune anoshanda ML uye DS masisitimu mukugadzira. Sarudzo dzinoitwa chete zvichibva pane data. Dhipatimendi rekugadzirisa data uye sainzi yedhipatimendi ndeimwe yeakakosha mukambani.
Zvinenge zvisingaite kuzadzisa zvese zviri pamusoro nedhipatimendi revanhu 2-3. Ini ndinofunga kuti ongororo iyi mhedzisiro marwadzo ari kukura zvishoma - vakomana havasati vaine munhu wekuzvienzanisa naye kuti vaone nhanho yavo zvine mwero.
Sezvinotarisirwa, masayendisiti edatha anopedza yakawanda yenguva yavo kwete pane super yakaoma masvomhu kana engineering, asi pakugadzirisa, kurodha, uye kuchenesa data. Muhunyanzvi hwese tinoona preprocessing mune yepamusoro 3. Asi isu kashoma kuona zvinhu zvakaoma sekugadzira ML modhi kana kushanda neBig Data mupamusoro 3 - chete pakati peML neDWH mainjiniya.
Kune zvakare mbiri mbiri dzinosiririsa. Nyanzvi dzinogadza 40% yemabasa avo pachavo. MuKazakhstan, kusvika parizvino chete makambani epamusoro eunicorn akaedza mabhenefiti ekushanda nedata hombe uye akadzidza maitiro ekuzviita zvine hunyanzvi. Vanoparidzira kumusika kuti Big Data uye Machine Learning inotonhorera, uye yechipiri echelon inotevera shure, asi hainzwisisi nguva dzose kuti kushanda nedata kunoshanda sei. Naizvozvo, tinoona kuti nyanzvi dzinozvigadzirira mabasa, uye mabhizinesi haagare achiziva zvavanoda.
Ndakashamisika kuti 20% yenyanzvi havatomboziva kana kambani yavo ine Data Warehouse. Hongu, uye nemasisitimu ekuchengetedza dhatabhesi hazvisi zvese zvakanaka - 41% shandisa MySQL, uye imwe 34% inoshandisa PostgreSQL. Izvi zvingarevei? Ivo vanoshanda pane nedata diki.
Mumubvunzo wezvekuchengetedza masisitimu, tinoona zvakare MySQL uye kunyange (!) Excel. Asi izvi zvinogona kuratidza, semuenzaniso, kuti makambani mazhinji haasati ave nechikumbiro chekushanda nedata hombe.
Pano zvinhu zvose zvakare hazvina kujeka. Kazhinji, mihoro yakanga iri shoma pane yandaitarisira.
Ini pachangu, zvakandiomera kuti ndifungidzire mainjiniya eML akagadzirira kushandira zviuru mazana maviri zvetenge - anogona kunge ari mudzidzi. Pamwe kugona kwenyanzvi dzakadai hakuna kusimba, kana zvichiri kuomera makambani kuti aongorore zvakakwana basa reData Science. Asi pamwe izvi zvinoratidza zvakare kuti musika uchiri pakutanga kwekukura kwayo. Uye nekufamba kwenguva, mwero wemihoro uchasimbiswa padanho rakaringana.
Source: www.habr.com