Mashandisiro atinoita Markov cheni mukuongorora mhinduro uye kutsvaga tsikidzi. Ne Python script

Zvakakosha kuti isu tinzwisise zvinoitika kuvadzidzi vedu panguva yekudzidziswa uye kuti zviitiko izvi zvinokanganisa sei mhedzisiro, saka tinovaka Mepu Yerwendo rweMutengi - mepu yeruzivo rwevatengi. Mushure mezvose, maitiro ekudzidza haasi chinhu chinoenderera uye chakakosha, iketani yezviitiko zvakabatana uye zviito zvemudzidzi, uye zviito izvi zvinogona kusiyana zvikuru pakati pevadzidzi vakasiyana. Iye zvino apedza chidzidzo chake: achazoitei zvakare? Ichaenda kubasa remumba here? Ichavhura here mobile application? Achachinja nzira here, okumbira kuchinja vadzidzisi? Uchananga kuchidzidzo chinotevera here? Kana kuti achabva aora mwoyo here? Zvinokwanisika here, nekuongorora iyi mepu, kuona mapatani anotungamirira kukubudirira kupedzisa kosi kana, zvakasiyana, kune "kudonha" kwemudzidzi?

Kazhinji, nyanzvi, inodhura yakavharwa-sosi maturusi anoshandiswa kuvaka CJM. Asi isu taida kuunza chimwe chinhu chiri nyore, chinoda kushoma kushanda nesimba uye, kana zvichibvira, yakavhurika sosi. Saka zano rakauya rekushandisa Markov cheni - uye takabudirira. Isu takavaka mepu, takadudzira data pamusoro pemaitiro emudzidzi muchimiro chegirafu, takaona mhinduro dzisiri pachena kumatambudziko ebhizinesi repasi rose, uye takatowana zvakavanzwa tsikidzi. Takaita zvese izvi tichishandisa yakavhurika sosi Python script mhinduro. Muchikamu chino ini ndichataura pamusoro pezviitiko zviviri neaya asinganyatsooneki mhedzisiro uye kugovera script nemunhu wese.

Saka, Makov maketani anoratidza mukana wekuchinja pakati pezviitiko. Heino muenzaniso wekutanga kubva kuWikipedia:

Pano "E" na "A" zviitiko, miseve ishanduko pakati padzo (kusanganisira kushanduka kubva pachiitiko kuenda kune chimwe chete), uye huremu hwemiseve ndiyo mukana wekuchinja ("weighted directed graph").

Wakashandisa chii?

Dunhu rakadzidziswa neyakajairwa Python mashandiro, ayo akadyiswa nematanda ezviitiko zvevadzidzi. Iyo girafu pane yakaguma matrix yakavakwa neNetworkX raibhurari.

Iyo log inotaridzika seizvi:

Iri ifaira recsv rine tafura yemakoramu matatu: mudzidzi id, zita rechiitiko, nguva yakaitika. Iyi minda mitatu yakakwana kuronda mafambiro emutengi, kuvaka mepu uye pakupedzisira kuwana Markov cheni.

Raibhurari inodzosa magirafu akagadzirwa ari .dot kana .gexf fomati. Kuti uone yekare, unogona kushandisa mahara Graphviz package (gvedit tool), takashanda nayo .gexf neGephi, zvakare yemahara.

Tevere ndinoda kupa mienzaniso miviri yekushandisa Markov cheni, iyo yakatibvumira kuti titarise patsva pazvinangwa zvedu, maitiro edzidzo, uye Skyeng ecosystem pachayo. Zvakanaka, gadzirisa zviputi.

Chekutanga nyaya: mobile application

Chekutanga, takaongorora rwendo rwevadzidzi kuburikidza nechigadzirwa chedu chinonyanya kufarirwa — kosi yeGeneral. Panguva iyoyo, ndaishanda mudhipatimendi revana reSkyeng uye taida kuona kuti nharembozha yakashanda sei nevateereri vevana vedu.

Kutora matanda nekuamhanyisa kuburikidza nescript, ndine chimwe chinhu chakadai:

Node yekutanga iStart General, uye pazasi pane matatu ekubuda mamodhi: mudzidzi "akarara," akachinja nzira, uye akapedza kosi.

• Akarara, "Kurara" - zvinoreva kuti haasisiri kutora zvidzidzo, zvichida akadonha. Tine tariro tinodaidza iyi nyika "yakarara", nekuti ... mupfungwa, achiri nemukana wekuenderera mberi nezvidzidzo zvake. Mugumisiro wakaipisisa kwatiri.
• Yakadonhedzwa general, Yakachinja kosi - yakachinja kubva kuGeneral kuenda kune chimwe chinhu uye yakarasika kune yedu Markov cheni.
• Ndapedza kosi, Yapedza kosi - yakanaka mamiriro, munhu apedza 80% yezvidzidzo (kwete zvidzidzo zvese zvinodiwa).

Kupinda mukirasi inobudirira node zvinoreva kubudirira kupedza chidzidzo pachikuva chedu pamwe chete nemudzidzisi. Inorekodha kufambira mberi kwekosi uye kusvika kune yaunoda mhedzisiro - "Yakapedza kosi." Zvakakosha kwatiri kuti vadzidzi vapinde zvakanyanya sezvinobvira.

Kuti uwane mhedziso yehuwandu hwehuwandu hweiyo nharembozha (app session node), takavaka maketani akaparadzana kune yega yega node dzekupedzisira tobva taenzanisa huremu hwemupendero paviri:

• kubva kuchikamu cheapp kudzokera kwairi;
• kubva pachikamu cheapp kuenda kukirasi yakabudirira;
• kubva kukirasi yakabudirira kuenda kuchikamu cheapp.

Kuruboshwe kune vadzidzi vakapedza kosi, kurudyi kune avo “vakarara”

Aya mapendero matatu anoratidza hukama pakati pekubudirira kwemudzidzi uye kushandisa kwavo mobile app. Taitarisira kuona kuti vadzidzi vanenge vapedza kosi vanenge vaine hukama hwakasimba neapplication pane vadzidzi vakarara. Nekudaro, muchokwadi isu takawana zvakapesana mhedzisiro:

• isu takava nechokwadi chekuti mapoka akasiyana evashandisi anodyidzana neyo mobile application zvakasiyana;
• vadzidzi vakabudirira vanoshandisa mobile application zvishoma zvakanyanya;
• vadzidzi vanorara vanoshandisa iyo mobile application zvakanyanya.

Izvi zvinoreva kuti vadzidzi vanorara vanotanga kupedza nguva yakawanda uye yakawanda mufoni application uye, pakupedzisira, vanoramba vari mairi nekusingaperi.

Pakutanga takashamiswa, asi pashure pokufunga nezvazvo, takaona kuti ikoku kwakanga kuri mugumisiro womuzvarirwo chose chose. Pane imwe nguva, ndakadzidza chiFrench ndega ndichishandisa maturusi maviri: nharembozha uye girama hurukuro paYouTube. Pakutanga, ndakaparadzanisa nguva pakati pavo muchiyero che 50 kusvika 50. Asi kushandiswa kunonyanya kunakidza, kune gamification, zvinhu zvose zviri nyore, zvinokurumidza uye zvakajeka, asi muhurukuro iwe unofanirwa kuongorora mazviri, nyora chimwe chinhu pasi. , dzidzira mubhuku rekunyorera. Zvishoma nezvishoma, ndakatanga kushandisa nguva yakawanda pa smartphone yangu, kusvikira mugove wayo wakura kusvika ku100%: kana ukapedza maawa matatu pairi, iwe unogadzira kunzwa kwenhema kwebasa rakapedzwa, nekuda kwekuti iwe hauna chido chekuenda uye kuteerera chero chinhu. .

Asi izvi zvingaitika sei? Mushure mezvose, isu takanyanya kugadzira mobile application, yakavakirwa mairi Ebbinghaus curve, akachigashira, chakaita kuti chitaridzike kuti vanhu vapedze nguva varimo, asi zvinoonekwa kuti chinovavhiringidza chete? Muchokwadi, chikonzero ndechekuti nharembozha application timu yakabata nemabasa ayo zvakanyanya, nekudaro yakave inotonhorera, inozvikwanira chigadzirwa ndokutanga kudonha kubva mu ecosystem yedu.

Nekuda kwetsvagiridzo, zvakava pachena kuti nharembozha yaida kuchinjwa neimwe nzira kuitira kuti isanyanya kuvhiringidza kubva pane chikuru kosi yekudzidza. Uye vose vana nevakuru. Basa iri riri kuenderera mberi.

Chechipiri nyaya: onboarding bugs

Onboarding inzira yekuwedzera yesarudzo paunenge uchinyoresa mudzidzi mutsva, uchibvisa matambudziko angangoitika ehunyanzvi mune ramangwana. Mamiriro ezvinhu ekutanga anofungidzira kuti munhu akanyoresa papeji yekumhara, akawana mukana kuakaundi yake yega, anobatikidzwa uye anopiwa chidzidzo chekutanga. Panguva imwecheteyo, tinocherekedza chikamu chikuru chematambudziko ehunyanzvi panguva yechidzidzo chekutanga: iyo isiriyo vhezheni yebrowser, maikorofoni kana ruzha haashande, mudzidzisi haagone kukurumidza kupa mhinduro, uye zvese izvi zvakanyanya kuoma kana zvasvika. kuvana. Naizvozvo, isu takagadzira imwezve application muakaundi yako yega, kwaunogona kupedzisa matanho mana akareruka: tarisa browser yako, kamera, maikorofoni uye simbisa kuti vabereki vachange vari padyo panguva yechidzidzo chekutanga (mushure mezvose, ndivo vanobhadhara. dzidzo yevana vavo).

Aya mapeji mashoma ekukwira airatidza faneli seizvi:

1: kutanga block nematatu akasiyana zvishoma (zvichienderana nemutengi) mafomu ekupinda uye password ekupinda.
2: bhokisi rekutarisa richibvumirana kune imwe nzira yekukwira.
2.1-2.3: Tarisa kuvepo kwevabereki, Chrome vhezheni uye ruzha.
3: bhuroka yekupedzisira.

Inotaridzika chaizvo: mumatanho maviri ekutanga, vazhinji vevashanyi vanobva, vachiziva kuti pane chimwe chinhu chekuzadza, tarisa, asi hapana nguva. Kana mutengi asvika padanho rechitatu, saka anenge atosvika pakupedzisira. Iko hakuna chikonzero chimwe chete chekufungidzira chero chinhu pane funnel.

Zvakadaro, isu takasarudza kuongorora kukwira kwedu kwete pane yakasarudzika-dimensional funnel, asi tichishandisa Markov cheni. Takabatidza zvimwe zviitiko zvishoma, takamhanyisa script uye tikawana izvi:

Mumhirizhonga iyi, chinhu chimwe chete chinogona kunzwisiswa zvakajeka: chimwe chinhu chakashata. Iyo onboarding maitiro ane mutsara, izvi zviri mukati medhizaini, hapafanirwe kuve newebhu yakadai yekubatanidza mairi. Uye pano zviri pachena pakarepo kuti mushandisi anokandwa pakati pematanho, pakati payo hapafanirwe kuve nekushanduka zvachose.

Panogona kunge paine zvikonzero zviviri zvemufananidzo uyu unoshamisa:

• madhiri akapinda mudura regi;
• Pane zvikanganiso muchigadzirwa pachacho - onboarding.

Chikonzero chekutanga chingangove chechokwadi, asi kuiyedza ibasa-yakanyanya, uye kugadzirisa matanda hakubatsire kuvandudza UX. Asi neyechipiri, kana iripo, chimwe chinhu chaifanira kuitwa nokukurumidza. Nokudaro, takaenda kunotarisa nodes, kuona mipendero isingafaniri kuvapo, uye kutsvaga zvikonzero zvekuitika kwavo. Takaona kuti vamwe vashandisi vakanamira uye vakafamba mudenderedzwa, vamwe vakadonha kubva pakati kusvika pekutanga, uye vamwe, musimboti, vakatadza kubuda pamatanho maviri ekutanga. Isu takaendesa iyo data kuQA - uye hongu, zvakazoitika kuti pakanga paine mabhugi akakwana mukukwira: ichi chakadaro-chigadzirwa, chidiki chemudonzvo, chisina kuedzwa zvakadzama, nekuti ... Hatina kutarisira matambudziko. Iye zvino maitiro ese ekurekodha achinja.

Iyi nyaya yakatiratidza kushandiswa kusingatarisirwi kweMarkov cheni mumunda weQA.

Edza iwe pachako!

Ndakaisa yangu Python script yekudzidzisa Markov maketani munzvimbo yeruzhinji - shandisa kune hutano hwako. Zvinyorwa paGitHub, mibvunzo inogona kubvunzwa pano, ini ndichaedza kupindura zvese.

Zvakanaka, zvinongedzo zvinobatsira: NetworkX raibhurari, Graphviz visualizer. Uye pano pane chinyorwa chaHabré nezve Markov cheni. Iwo magirafu ari muchinyorwa anogadzirwa uchishandisa Gephi.

Source: www.habr.com