Chinyorwa chinokurukura nzvimbo dzemashandisirwo enguva dzakatevedzana, matambudziko anofanirwa kugadziriswa, uye maalgorithms anoshandiswa. Kufanotaura kwenguva kunoshandiswa mumabasa akadai sekufanotaura kudiwa, nzvimbo yekubatanidza mutoro, mugwagwa uye Internet traffic, kugadzirisa dambudziko rekutanga rinotonhora mumasisitimu ekurumbidza, uye kutsvaga anomalies mumaitiro emidziyo nevashandisi.
Ngatitarisei mabasa mune zvakadzama.
1) Kuda kufanotaura.
Chinangwa: kuderedza mitengo yekuchengetera zvinhu uye kugadzirisa zvirongwa zvebasa revashandi.
Maitiro ekuzvigadzirisa: kuva nefungidziro yekutenga kwezvinhu uye nhamba yevatengi, isu tinodzikisa huwandu hwezvinhu mudura uye tinochengeta chaizvo izvo zvichatengwa munguva yakatarwa. Kuziva nhamba yevatengi chero nguva ipi zvayo, tichagadzira hurongwa hwekushanda hwakakwana kuitira kuti pave nenhamba yakakwana yevashandi vane mari shoma.
2) Kufanotaura mutoro pabasa rekutumira
Chinangwa: kudzivirira Logistics kudonha panguva yepamusoro mitoro.
Maitiro ekuzvigadzirisa: kufanotaura huwandu hwemirairo, hunza huwandu hwakakwana hwemota uye vatakuri pamutsetse.
3) Kufanotaura mutoro pane yekuonana nzvimbo
Chinangwa: kuve nechokwadi chekuwanikwa kunodiwa kwenzvimbo yekuonana uku uchidzikisa mari yemuhoro.
Maitiro ekugadzirisa: kufanotaura huwandu hwekufona nekufamba kwenguva, kugadzira iyo yakakwana hurongwa hwevashandisi.
4) Kufanotaura kwemigwagwa
Chinangwa: kufanotaura nhamba yemaseva uye bandwidth yekushanda kwakagadzikana. Kuti sevhisi yako isaparare pazuva rekutanga kweyakakurumbira TV akatevedzana kana nhabvu mutambo 😉
5) Kufanotaura nguva yakakwana yekuunganidza ATM
Chinangwa: kuderedza huwandu hwemari yakachengetwa muATM network
6) Mhinduro kune kutonhora kwekutanga dambudziko mumasitimu ekurudziro
Chinangwa: Kurudzira zvigadzirwa zvinoenderana kune vashandisi vatsva.
Kana mushandisi atenga akati wandei, yakabatana kusefa algorithm inogona kuvakwa yekurudziro, asi kana pasina ruzivo nezve mushandisi, zvakaringana kurumbidza zvigadzirwa zvakakurumbira.
Solution: Kuzivikanwa kwezvigadzirwa zvinoenderana nenguva iyo kurudziro yakaitwa. Kushandisa fungidziro yenguva kunobatsira kuona zvigadzirwa zvakakodzera pane chero nguva yakapihwa panguva.
Takatarisa ma hacks ehupenyu ekuvaka anorumbidza masisitimu mukati .
7) Tsvaga anomalies
Chinangwa: kuona matambudziko mukushanda kwemidziyo uye isiri-yakajairwa mamiriro mubhizinesi
Solution: Kana kukosha kwakayerwa kuri kunze kweiyo fungidziro yenguva yekuvimba, kusanzwisisika kwaonekwa. Kana ichi chiri chigadzirwa chemagetsi enyukireya, inguva yekuwedzera sikweya yechinhambwe 😉
Algorithms yekugadzirisa dambudziko
1) Kufamba pakati
Iyo yakapfava algorithm ndiyo inofamba pakati. Ngativerengei kukosha kweavhareji pazvinhu zvishoma zvekupedzisira toita kufanotaura. Kune mamiriro ekunze akareba kupfuura mazuva gumi, nzira yakafanana inoshandiswa.
Kana zvakakosha kuti hunhu hwekupedzisira munhevedzano huwedzere huremu, tinosuma coefficients zvichienderana nedaro rezuva, tichiwana huremu modhi:
Saka, iwe unogona kuseta W coefficient kuitira kuti huremu hwakanyanya hudonhe pamazuva maviri ekupedzisira uye nemazuva ekupinda.
Tichifunga nezve cyclical factors
Hunhu hwekurudziro hunogona kukanganiswa nezvinhu zvinotenderera, senge sangana nezuva revhiki, zuva, mazororo apfuura, nezvimwe.
Mupunga. 1. Muenzaniso wenguva yakatevedzana kuora kuita maitiro, mwaka wechikamu uye ruzha
Exponential smoothing ndiyo mhinduro yekufunga nezve cyclical factor.
Ngatitarisei nzira nhatu dzekutanga
1. Kutsvedzerera kuri nyore (Brown modhi)
Inomiririra kuverenga kweavhareji yakayerwa pazvikamu zviviri zvekupedzisira zvenhevedzano.
2. Kutsvedzerera kaviri (Holt modhi)
Inotora muakaunti shanduko yemaitiro uye kuchinjika mune zvakasara zvimiro zvakatenderedza maitiro aya.
Isu tinoverenga kufanotaura kwekuchinja mune zvakasara ® uye maitiro (d). Kukosha kwekupedzisira kwe y ndiko kuwanda kwezviwandu zviviri izvi.
3. Kutsetseka katatu (Holt-Winters modhi)
Katatu kutsvedzerera kunowedzerawo kufunga nezvekusiyana kwemwaka.
Mafomula ekutsvedzerera katatu.
ARIMA uye SARIMA algorithm
Iyo yakasarudzika yenguva yakatevedzana yekushandiswa kweARIMA ndiko kubatana pakati peakapfuura kukosha kwakabatana neazvino uye anotevera.
SARIMA - yekuwedzera kune yakatevedzana ine mwaka wechikamu. SARIMAX ndeyekuwedzerwa kunosanganisira chikamu chekunze chekugadzirisa.
ARIMA modhi inobvumidza iwe kutevedzera yakabatanidzwa kana mutsauko-yakamira nguva yakatevedzana.
Iyo ARIMA nzira yekuteedzera nguva ndeyekuti kumira kweiyo nhevedzano kunotanga kuongororwa.
Tevere, iyo nhevedzano inoshandurwa nekutora mutsauko weiyo yakakodzera kurongeka, uye ARMA modhi inogadzirirwa iyo yakashandurwa modhi.
ARMA mutsara wakawanda wekudzoreredza modhi.
Zvakakosha kuti mutsara umire, i.e. kureva uye kusiyana hakuna kuchinja. Kana iyo nhevedzano isiri-yakamira, inofanirwa kuunzwa kune yakamira fomu.
XGBoost - taizovepi pasina iyo?
Kana nhevedzano isina yemukati yakaratidza chimiro, asi paine zvekunze zvinopesvedzera zvinhu (maneja, mamiriro ekunze, nezvimwewo), saka unogona kushandisa zvakachengeteka mhando dzekudzidza dzemuchina sekusimudzira, masango asina kurongeka, kudzoreredza, neural network uye SVM.
Kubva pane chiitiko chechikwata DATA4, fungidziro yenguva, rimwe remabasa makuru ekugadzirisa optimization yemitengo yekuchengetera, mutengo wevashandi, kugadzirisa kuchengetedza kweATM network, logistics uye masisitimu ekukurudzira kuvaka. Mhando dzakaomarara seSARIMA dzinopa mhedzisiro yemhando yepamusoro, asi dzinotora nguva uye dzinongokodzera mamwe mabasa.
Muchinyorwa chinotevera tichatarisa nzira huru dzekutsvaga anomalies.
Kuti uve nechokwadi chekuti zvinyorwa zvinoenderana nezvako zvaunofarira, tora ongororo iri pazasi, kana nyora mune zvakataurwa kuti ndedzipi misoro yekunyora nezvayo mune zvinotevera zvinyorwa.
Vashandisi vakanyoresa chete ndivo vanogona kutora chikamu muongororo. , Munogamuchirwa.
Zvinyorwa zvenyaya ipi yaunofarira?
-
Recommender masisitimu
-
Kuzivikanwa kwemufananidzo
-
Kutaura uye kunyora zvinyorwa
-
Zvivakwa zvitsva muDNN
-
Временные ряды и поиск аномалий
-
ML mubhizinesi, shandisa makesi
17 vashandisi vakavhota. 3 vashandisi vakaramba.
Source: www.habr.com