Kutanga chirongwa chitsva chekudzidza muchina kunounza kukurumidza kwechido, uye zvingave zvinoyedza kusvetukira kumucheto wakadzika. Kune akawanda emazuva ano ekucheka-kumucheto modhi kana yakaoma algorithms yaungave wakaverenga nezvayo. Vanovimbisa mibairo inokatyamadza, uye kudzivisa muedzo wekuedza navo ipapo ipapo ibasa rakaoma.  Chero muzvinabhizimusi wemazuva ano anoda kuyedza matekiniki emazuva ano uye kuratidza mapurojekiti akaomesesa (uye akabudirira) kunharaunda. Asi, kushingairira uku, kunyangwe kwakanaka, dzimwe nguva kunogona kutora nguva yakakosha paunenge uchikwenenzvera hyperparameter uye kusangana nekuoma kwekushandisa maitiro akaomarara.  Mukuita uku, pane mubvunzo mumwechete mukuru unoda kubvunzwa:  Isu tinoyera sei chaizvo kushanda kwemuenzaniso wedu?  Kutsvaga kuti kuomarara kwemodhi yedu kwakarurama here kana kuti kuita kwacho kuri nani zvechokwadi kunogona kunetsa. Izvi zvinoitika kana   Pano, kuva nemuenzaniso wekutanga kunove kwakakosha zvikuru. Nheyo yekutanga inopa iyo yakakosha yereferensi poindi - yakatwasuka, inokurumidza kuvaka, uye nemasikirwo anotsanangurwa. Zvinoshamisa kuti kazhinji muenzaniso wekutanga, unogona kungotora 10% yekuedza kwese kwekusimudzira, unogona kuwana kusvika ku90% yemaitiro anodiwa, ichigadzira nzira inoshanda kune mhedzisiro inonzwisisika. pasina chirevo chakareruka chekutarisa.  Pfungwa yekutanga yakapusa haingori nzira iri nyore kune vanotanga - itsika yakakosha inogara yakakosha pamatanho ese ebasa resainzi yedata. Iyo nzira yekumisa uye chiyeuchidzo chikuru chekuenzanisa chishuvo chedu chekuomarara nezvinoita zvezvakajeka, zviri nyore kunzwisisa, uye zvinogoneka mhinduro.  Kunzwisisa Baseline Models  A baseline modhi ndiyo   inoshandiswa kugadzirisa dambudziko. Kazhinji, aya mamodheru anosanganisira mutsara kudzokororwa kwezvinoenderera mhedzisiro kana logistic regression kune categorical mhedzisiro. Semuenzaniso, kudzoreredza kwemutsara kunogona kufanotaura kudzoserwa kwemasheya zvichienderana nenhoroondo yemutengo wenhoroondo, nepo kudzoreredza kwemaitiro kunogona kuronga vanonyorera kiredhiti senjodzi yakanyanya kana yakaderera. inonyanya kukosha shanduro  Iyi nzira inosiyana nemhando dzakaomesesa senge neural network kana ensemble nzira, iyo, kunyange ine simba, inogona kuita kuti kubata dambudziko kunyanye kuoma uye kuwedzera nguva inodiwa yebudiriro nekuda kwekuoma kwavo uye kwakakosha zviwanikwa zvemakomputa.   Zvakanakira Kutanga neBaseline Model   Benchmarking  Benchmarking inhanho yakakosha yekutanga mukuvandudza chero modhi yeML. Kana iwe ukamisa yekutanga modhi, iwe unomisikidza yakakosha metric yekuita iyo ese mamodheru anouya mushure (ayo anowanzo kuve akaomarara) anofanirwa kupfuura kururamisa kuoma kwavo uye kushandisa zviwanikwa. Iyi nzira haingori cheki huru yehutsanana chete asiwo inosimbisa tarisiro yako uye inokupa chiyero chakajeka chekufambira mberi.  Semuenzaniso, fungidzira uchigadzira modhi yekufanotaura mafambiro emusika wemari uchishandisa yakapusa yekufamba pakati (SMA) seyokutanga. Iyi SMA inogona kushandisa yenguva pfupi yenhoroondo data kufanotaura mitengo yenguva yemberi, kuwana huchokwadi hwekutanga hwe60% mukufanotaura mafambiro emusika nenzira kwayo. Iyi modhi inozoisa bhenji kune chero mhando dzepamberi dzinotevera. Kana muenzaniso wakaoma, wakadai seyetiweki Yenguva Yakareba Yenguva Yekurangarira (LSTM), ikazogadzirwa gare gare uye ikawana huchokwadi hwe65%, kuwedzera kwekuita kunogona kuyerwa nemazvo pakatarisana nekutanga 60% yekutanga.  Kuenzanisa uku kwakakosha pakuona kana kuvandudzwa kwe5% mukurongeka kunopembedza kuwedzera kuomarara uye computational zvinodiwa zveLSTM. Pasina hwaro hwekutanga seizvi, kuita sarudzo dzine ruzivo nezve scalability uye inoshanda mamodheru akaomarara zvinova zvinonetsa.  Iyi nzira yekuisa mabhenji inova nechokwadi chekuti kuvandudzwa kwemuenzaniso kuoma kwakakodzera uye kunozoguma nekuvandudzwa chaiko, zvese izvi zvichiita kuti hurongwa hwekuvandudza huenderane nemhedzisiro inobudirira.   Mutengo-Kubudirira  Kutevera nzira inodhura muML ndiyo yakakosha. Kunyanya kana   Paunotanga neyokutanga modhi, unoderedza zviwanikwa uye nguva inodiwa yekutanga modhi yekuvandudza uye kuyedzwa. Izvi zvinoreva kukurumidza prototyping - uye izvo zvakakosha pamhinduro yekukurumidza nekuvandudzwa. iwe uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge uchinge rwakaita.  Neiyi yekutanga, chero kuomarara kwaunowedzera ikozvino kunogona kunyatsoongororwa.  Semuenzaniso, kana iwe uchida kuita shanduko kuenda kune yakanyanya kuomarara algorithm senge vector autoregression (VAR) uye woona kuti inongowedzera zvishoma kufembera, iwe unofanirwa kufunga patsva kana iyi kugadziridzwa kudiki kuchiruramisa izvo zvekuwedzera computational zvinodiwa uye kuoma. Mhinduro ingava kwete. Ipapo iyo iri nyore modhi inoramba iri yakanyanya kudhura-inoshanda sarudzo.  Nekutarisa pakudhura-kuita, unoita shuwa kuti zviwanikwa zvinoshandiswa nemazvo uye kuwana zvinopfuura kungowedzera tekinoroji. Zvakare, inopa inoshanda, yakakosha-yakawedzerwa mhinduro dzinoruramiswa maererano nekuvandudzwa kwekuita uye kugoverwa kwezviwanikwa. Nenzira iyi, kudyara kwega kwega mukuoma kwemuenzaniso kunotenderwa, izvo zvinopa kune zvinangwa zvepurojekiti zvakazara pasina mari inodhura.   Kujeka uye Kududzira  Muzvikamu zvakaita sezvemari umo sarudzo dzinofanirwa kutevedzera zvakaomesesa zvimiro, kuve pachena kwemamodheru haingori mukana webhizinesi. Iyo inzira yehungwaru inobatsira zvakanyanya mukuita kwekusangana mirau uye kufambisa kutaurirana kuri nyore nevanobata vanogona kunge vasina (yakadzama) yehunyanzvi kumashure.  Ngatitorei modhi yedu yeSMA. Inodudzirwa zviri nyore nekuti mibairo yayo yakanangana nedata rekuisa. Izvi zvinoita kuti zvive nyore kutsanangura kuti kupinza kwega kwega kunopesvedzera sei zvakafanotaurwa. Kana sarudzo dzakavakirwa pakufanotaura kwemuenzaniso dzinoda kururamiswa kune vekunze vanodzora kana mukati kune vasiri tekinoroji nhengo dzechikwata, kupusa uku ndiko kwakakosha kune ako maitiro.  Kana sarudzo yakavakirwa pakufanotaura kweSMA modhi ikabvunzwa, kujeka kweiyo modhi kunobvumira kutsanangura nekukurumidza uye kwakapfava kweiyo logic kumashure kwebasa rayo. Izvi zvinogona kubatsira nekuongorora kwemitemo uye ongororo uye kugadzirisa kuvimba uye kugamuchirwa pakati pevashandisi nevanoita sarudzo. Uyezve, sezvo kuomarara kwemuenzaniso kunowedzera, semuenzaniso uchienda kune akaomesesa algorithms seARIMA kana VAR modhi kune mamwe mafungidziro asina kujeka, kududzira kweiyo yekutanga SMA baseline inova bhenji kune iyo nhanho yetsananguro yaunoda kuratidza.  Nekushandisa regressors senge chimiro chakakosha zvibodzwa kana SHAP makoshero akasanganiswa nemhando dzakaomarara, kufambira mberi kweimwe imwe modhi kuita kunoramba kuri pachena. Izvi zvinobatsira chinangwa chekuchengetedza nzira kuti dzisaraswa kune mamwe emhando yepamusoro. Pfungwa yeiyo yakapfava yekutanga modhi ndeyekugara ichiita iyo mamiriro ekuti chimiro chese uye kukosha kuchachengetwa kunyangwe huwandu hwekuoma hunowedzera. Izvi zvinovimbisa zviga zvekuteedzera uye kutaurirana kunozoshanda.   Risk Management  Kutarisira njodzi ndechimwe chinhu chakakosha pakugadzira mhando dzemakina ekudzidza, kunyanya muzvikamu zvakaita sezvemari uko kufanotaura kwakaringana uye kwakavimbika kune chekuita mukuita sarudzo. Kuve neyakapfava yekutanga modhi inzira huru yekugadzirisa njodzi idzi.  Iyo yakatwasuka yekutanga inopa inonzwisisika yekutanga nzvimbo, iyo inokutendera iwe kuti zvishoma nezvishoma (uye zvakachengeteka) uwedzere zvigadziriso kune modhi yakaoma.  Semuenzaniso, iyo SMA modhi (ichiri yakakosha) inoita hwaro hwakasimba hwekutsvaga pasi pemaitiro uye zvingangove zvisizvo mukufamba kwemutengo wemasheya. Kuishandisa kunobatsira kuona zviratidzo zvekutanga zvekusadzikama kana kuti maitiro emusika asina kunaka. Kuita izvo kwakakosha, kudzivirira njodzi dzakakura dzemari usati watumira mamwe akaomarara ekufungidzira algorithms.  Uyezve, kushandisa yekutanga modhi kunoderedza njodzi yekuwedzeredza. Iyo igomba rinowanzoitika mukuenzanisira kwemari. Kuwedzeredza kunoitika kana modhi yakanyatsokwenenzverwa kune data renhoroondo uye inotora ruzha pane iyo iri pasi pepateni. Nekuda kweizvi, iwe unogona kuwana fungidziro dzinotsausa uye kuwana asina kuvimbika ekutengesa nzira semhedzisiro. Modhi yakapfava ine mashoma ma paramita hainyanyi kutarisana nenyaya iyi, kuve nechokwadi chekuti fungidziro yainopa inowanzoshanda kune isingaonekwe data.  Kuwedzera kuomarara sezvo SMA inofambira mberi pane diki inofamba avhareji modhi seARIMA neVAR inova yakaomarara, chimiro chakareruka cheSMA chinogona kutibatsira kufunga zvine hungwaru kushanda kwega kwega kwakawedzera kuoma. Iyi nhanho nhanho yekuvandudza mukuoma kunobatsira kuchengetedza kutonga pamusoro pekuita kwemuenzaniso, kuve nechokwadi chekuti imwe neimwe yekuwedzera yakaomesesa layer inopa bhenefiti yakajeka uye haiunze njodzi isina kufanira.  Iyi yakarongeka nzira yekuwedzera modhi yakaoma inobatsira mukunzwisisa kuti shanduko kune iyo modhi inokanganisa maitiro ayo uye kuvimbika. Inovimbisawo kuti njodzi dzinogara dzakagadziriswa zvakanaka. Paunotanga neyakapfava yekutanga uye nekungwarira kudzora yega yega nhanho yekusimudzira, iwe unoita shuwa kuti mamodheru ekufungidzira anoramba aine simba uye akachengeteka, achitsigira kuita sarudzo yemari.   Mafungiro Akakosha Pakushandisa Baseline Models  Kuti usarudze yakanyatsokodzera yekutanga modhi, iwe unofanirwa kunzwisisa dambudziko rebhizinesi uye data data. Semuyenzaniso, nguva-yakatevedzana fungidziro yemisika yemari inogona kutanga neiyo ARIMA modhi senheyo yekutanga kutora simba renguva nenzira iri nyore. Hunhu hwedata uye preprocessing zvakare inotamba mabasa akakosha; kunyange iyo yakapfava modhi inogona kuita zvisina kunaka kana ichipihwa isina kukwana kana isina kunyatsogadziriswa data.  Uye chekupedzisira, kuziva nguva yekuchinja kubva kune yekutanga kuenda kune yakaoma modhi yakakosha. Iyi sarudzo inofanirwa kutungamirwa nekuwedzera kuyedzwa uye kusimbiswa, zvinoenderana neAgile's iterative maitiro.   To Sum Up  Kutanga mapurojekiti ako ekudzidza muchina nekuunza yakapfava yekutanga modhi haingori nhanho yekutanga. Izano. Iro zano rinoenderana neAgile nzira dzinosimudzira kushanda zvakanaka, kugona, uye kuchinjika. Kusvika purojekiti yako nenzira iyi kunogona kuwedzera zvakanyanya mhedzisiro yeprojekiti nekuona kuti kuwedzera kwese kwekuoma kwakakosheswa uye kunowedzera kukosha kunobatika. Kugamuchira zviri nyore chinhu chine simba. Iro izano rakakura kwazvo mundima semari umo sarudzo dzinofanirwa kukurumidza.

Arweave

Read My Stories

Odhiyo iyi inogadzirwa mumutauro wekutanga wenyaya!

Kurebesa; Kuverenga

Make resilience your competitive advantage

Kutanga Nyore: Iyo Strategic Advantage yeBaseline Models muMuchina Kudzidza

About Author

MACOMMENTS

HONGA TAGS

NYAYA IYI YAKAPIWA MUNA

Related Stories

Empowering Senior Professionals: Interview with Startups of the Year 2024 Nominee, SNATIKA

On the Position of Large Factories

Snowdrop & Other Tales by Jacob Grimm and Wilhelm Grimm - Table of Links

Meet SNATIKA, Winner of HackerNoon's Startups of The Year 2024 in Training And Consulting

Empowering Senior Professionals: Interview with Startups of the Year 2024 Nominee, SNATIKA

On the Position of Large Factories

Snowdrop & Other Tales by Jacob Grimm and Wilhelm Grimm - Table of Links

Meet SNATIKA, Winner of HackerNoon's Startups of The Year 2024 in Training And Consulting

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps