Zviri Mukati[Viga][Ratidza]
Imwe yemaitiro ekutanga emhando ipi neipi yezviitwa zvemakambani mashandisirwo anobudirira eruzivo. Pane imwe nguva, vhoriyamu yedata yakagadzirwa inodarika kugona kwekutanga kugadzirisa.
Ndipo panopinda muchina kudzidza algorithms. Zvisinei, izvi zvisati zvaitika, ruzivo runofanira kudzidzwa uye kududzirwa. Muchidimbu, ndiyo inoshandiswa kudzidza muchina usina kutariswa.
Muchinyorwa chino, isu tichaongorora zvakadzama zvisina kutariswa muchina kudzidza, kusanganisira maalgorithms ayo, makesi ekushandisa, uye zvimwe zvakawanda.
Chii chinonzi Unsupervised Machine Learning?
Kusatariswa kwemichina yekudzidza algorithms inotaridza mapatani mune dataset isina zvinozivikanwa kana kunyorwa mhedzisiro. Supervised muchina kudzidza algorithms iva nezvakaburitswa.
Kuziva musiyano uyu kunokubatsira kuti unzwisise kuti sei nzira dzisina kutariswa dzekudzidza muchina dzisingagone kushandiswa kugadzirisa kudzoreredza kana nyaya dzemhando, sezvo usingazive kukosha/mhinduro yedata rekubuda. Iwe haugone kudzidzisa algorithm kazhinji kana iwe usingazive kukosha / mhinduro.
Uyezve, kudzidza kusina kutariswa kunogona kushandiswa kuona chimiro chakakosha che data. Aya maalgorithms anoona mapatani akavanzika kana mapoka edata pasina kudiwa kwekudyidzana kwevanhu.
Kugona kwayo kuona kufanana uye kusiyanisa muruzivo kunoita kuti ive sarudzo yakanaka yekuongorora data, matekiniki ekutengesa-muchinjika, kupatsanurwa kwevatengi, uye kuzivikanwa kwemifananidzo.
Funga nezvechiitiko chinotevera: uri muchitoro chegirosari woona muchero usingazivikanwe wausati wamboona. Iwe unogona nyore kusiyanisa muchero usingazivikanwe wakasiyana nemimwe muchero wakatenderedza zvichienderana nekuona kwako chimiro, saizi, kana ruvara.
Usupervised Machine Kudzidza Algorithms
Clustering
Clustering pasina kupokana ndiyo inonyanya kushandiswa isina kutariswa nzira yekudzidza. Iyi nzira inoisa zvinhu zvine chekuita nedata mumasumbu anogadzirwa zvisina tsarukano.
Nechega, modhi yeML inoona chero mapatani, akafanana, uye/kana misiyano mune isina kukamurwa data chimiro. Muenzaniso unozokwanisa kuwana chero mapoka echisikigo kana makirasi mune data.
Types
Kune akati wandei maitiro ekubatanidza anogona kushandiswa. Ngatitarisei zvakanyanya kukosha kutanga.
- Exclusive clustering, dzimwe nguva inozivikanwa se "yakaoma" clustering, imhando yeboka umo chidimbu chimwe che data chiri chesumbu rimwe chete.
- Kubatanidza kubatanidza, kunowanzozivikanwa se "kunyorova" kuunganidza, kunobvumira zvinhu zvedata kuti zvive zveanopfuura sumbu rimwe kune madhigirii akasiyana. Uyezve, probabilistic clustering inogona kushandiswa kugadzirisa "zvakapfava" kusangana kana matambudziko ekufungidzira density, pamwe nekuongorora mukana kana mukana wemapoinzi edata emamwe masumbu.
- Kugadzira hierarchy yezvinhu zvakaunganidzwa zve data ndicho chinangwa chehierarchical clustering, sezita rinoratidza. Zvinhu zve data zvinogadziriswa kana kusanganiswa zvichibva pane hierarchy kugadzira masumbu.
Shandisa kesi:
- Anomaly Kuonekwa:
Chero mhando yekunze mune data inogona kuwonekwa uchishandisa clustering. Makambani mune zvekufambisa uye zvekutakura, semuenzaniso, anogona kushandisa kusarudzika kuona zvipingamupinyi zvekugadzirisa zvinhu kana kuburitsa zvikamu zvakakanganisika zvemuchina (predictive maintenance).
Masangano ezvemari anogona kushandisa tekinoroji kuona kutengeserana kwechitsotsi uye kupindura nekukurumidza, zvichigona kuchengetedza mari yakawanda. Dzidza zvakawanda nezve kuona zvisirizvo uye hutsotsi nekuona vhidhiyo yedu.
- Segmentation yevatengi uye misika:
Clustering algorithms inogona kubatsira mukuunganidza vanhu vane hunhu hwakafanana uye kugadzira vatengi vanhu kuti vawedzere kushanda kwekushambadzira uye kwakanangwa zvirongwa.
K-Means
K-zvinoreva inzira yekubatanidza iyo inozivikanwawo sekugovera kana kupatsanura. Inopatsanura mapoinzi edata kuita nhamba yakafanotemerwa yemasumbu anozivikanwa seK.
Mune iyo K-zvinoreva nzira, K ndiyo yekuisa sezvo iwe uchiudza komputa kuti mangani masumbu aunoda kuratidza mune yako data. Chinhu chega chega che data chinozopihwa kune yepedyo cluster centre, inozivikanwa secentroid (madotsi matema pamufananidzo).
Iyo yekupedzisira inoshanda senzvimbo dzekuchengetedza data. Iyo nzira yekubatanidza inogona kuitwa kakawanda kusvika masumbu anyatsotsanangurwa.
Fuzzy K-zvinoreva
Fuzzy K-zvinoreva kuwedzera kweK-nzira nzira, iyo inoshandiswa kuita kupindirana kwekubatana. Kusiyana neK-zvinoreva maitiro, akapusa K-zvinoreva zvinoratidza kuti data mapoinzi anogona kunge ari emasumbu mazhinji ane madhigirii akasiyana epedyo kune imwe neimwe.
Nhambwe iri pakati pemapoinzi edata necentroid yecluster inoshandiswa kuverenga kuswedera. Nekuda kweizvozvo, panogona kuve nezviitiko apo masumbu akasiyana anopindirana.
Gaussian Musanganiswa Models
Gaussian Mixture Models (GMMs) inzira inoshandiswa mukuunganidzira kubatanidza. Nekuti zvinoreva uye musiyano hazvizivikanwe, mamodheru anofungidzira kuti kune nhamba yakatarwa yekugovera kweGaussian, imwe neimwe ichimiririra sumbu rakasiyana.
Kuti uone kuti isumbu ripi re data point ndeye, iyo nzira inonyanya kushandiswa.
Hierarchical Clustering
Iyo hierarchical clustering strategy inogona kutanga neyega yega data point yakapihwa kune rakasiyana cluster. Iwo masumbu maviri ari pedyo nepedyo anobva asanganiswa kuita sumbu rimwe. Iterative kubatanidza inoenderera kusvikira sumbu rimwe chete rasara kumusoro.
Iyi nzira inozivikanwa sepasi-kumusoro kana agglomerative. Kana iwe ukatanga nezvinhu zvese zve data zvakasungirirwa kuchikwata chimwe chete uye wozoita kupatsanurwa kusvika chimwe nechimwe che data chapihwa sechikwata chakasiyana, iyo nzira inozivikanwa seyepamusoro-pasi kana inoparadzanisa hierarchical clustering.
Apriori Algorithm
Kuongorora kwemusika basket kwakawedzera apriori algorithms, zvichikonzera akasiyana einjini ekurudziro emapuratifomu emimhanzi uye zvitoro zvepamhepo.
Iwo anoshandiswa mumaseti ekutengesa kuti awane kazhinji zvinhu, kana zvikwata zvezvinhu, kuitira kufanotaura mukana wekushandisa chimwe chigadzirwa zvichienderana nekudyiwa kwechimwe.
Semuenzaniso, kana ndikatanga kuridza redhiyo yeOneRepublic paSpotify ine "Kuverenga Nyeredzi," imwe yedzimwe nziyo dziri pachiteshi ichi zvirokwazvo ichava rwiyo rweImagine Dragon, senge "Bad Liar."
Izvi zvinobva pamaitiro angu ekuteerera ekare pamwe chete nematererero evamwe. Nzira dzeApriori dzinoverenga zvinhu uchishandisa muti wehashi, uchiyambuka dhatabheti pahupamhi-kutanga.
Kuderedza Kuderedza
Kudzikiswa kwedimensionality imhando yekudzidza kusingatarisirwe kunoshandisa muunganidzwa wemazano kudzikisa huwandu hwezvimiro - kana zviyero - mudhatabheti. Titenderei kuti tijekese.
Zvinogona kuyedza kuisa data rakawanda sezvinobvira paunenge uchigadzira yako dataset yekudzidza muchina. Usatiite zvisizvo: zano iri rinoshanda nemazvo nekuti data rakawanda rinowanzo buritsa zvakawanikwa zvakarurama.
Fungidzira kuti data yakachengetwa muN-dimensional space, nechinhu chimwe nechimwe chinomiririra chiyero chakasiyana. Panogona kunge paine mazana ezviyero kana paine data rakawanda.
Funga nezveExcel maspredishiti, ane makoramu anomiririra maitiro nemitsara inomiririra data zvinhu. Kana paine zviyero zvakawandisa, ML algorithms inogona kuita zvisina kunaka uye kuona data zvinogona kuoma.
Saka zvinoita kuti zvive zvine musoro kudzikamisa hunhu kana hukuru, uye kupa ruzivo rwakakodzera chete. Dimensionality kuderedza ndizvo chete. Inobvumira huwandu hunogoneka hwekuisa data pasina kukanganisa kuperera kwedataset.
Chikuru Chikamu Chekuongorora (PCA)
Chikuru chechikamu chekuongorora inzira yekudzikisa dimensionality. Inoshandiswa kudzikisa huwandu hwezvimiro mumaseti makuru, zvichikonzera kureruka kwedata pasina kupira chokwadi.
Kudzvanywa kwedataset kunoitwa nenzira inozivikanwa sechinhu chekubvisa. Zvinotaridza kuti maelement kubva museti yepakutanga anosanganiswa kuita itsva, diki. Hunhu hutsva uhu hunozivikanwa sezvikamu zvekutanga.
Ehe, kune mamwe maalgorithms aunogona kushandisa mune ako asina kutariswa ekudzidza maapplication. Iwo akanyorwa pamusoro apa ndiwo anonyanya kupararira, ndosaka achikurukurwa zvakadzama.
Kushandiswa Kwekudzidza Kusina Kutariswa
- Nzira dzekudzidza dzisina kutariswa dzinoshandiswa pakuona mabasa ekuona sekuziva chinhu.
- Kudzidza kwemichina isina kutariswa kunopa zvinhu zvakakosha kumasisitimu ekufungidzira ekurapa, sekuzivikanwa kwemifananidzo, kupatsanura, uye kupatsanurwa, ayo anoshandiswa muradiology uye pathology kuongorora varwere nekukurumidza uye nekuvimbika.
- Kudzidzira kusingatarisirwe kunogona kubatsira kuona mafambiro e data anogona kushandiswa kugadzira nzira dzekutengesa dzinoshanda uchishandisa data rakapfuura pamaitiro evatengi. Munguva yekutarisa, izvi zvinoshandiswa nemabhizinesi epamhepo kupa zano rekuwedzera kune vatengi.
- Nzira dzekudzidza dzisina kutariswa dzinogona kusefa kuburikidza nemavhoriyamu makuru e data kuti uwane anoburitsa. Izvi zvisizvo zvinogona kusimudza chiziviso chemidziyo isingashande, kukanganisa kwevanhu, kana kutyorwa kwekuchengetedza.
Matambudziko ane Kusatarisirwa kudzidza
Kudzidzira kusingatarisirwe kunokwezva nenzira dzakasiyana siyana, kubva pakugona kuwana zvakakosha maonero data kuitira kudzivirira kudhura data label mabasa. Zvisinei, pane zvipingamupinyi zvakawanda pakushandisa nzira iyi kudzidzisa michina yekudzidza mamodheru kuti unofanira kuziva. Heino mimwe mienzaniso.
- Sezvo data yekupinda isina mavara anoshanda sekiyi dzekupindura, mhinduro dzisina kutariswa dzemhando dzekudzidza dzinogona kunge dzisina kunyatsojeka.
- Kudzidza kusingatarisirwe kunowanzo shanda nemaseti makuru, ayo anogona kuwedzera computational kuomarara.
- Iyo nzira inoda kusimbiswa kwakabuda nevanhu, vangave vemukati kana vekunze nyanzvi munyaya yekubvunza.
- Algorithms inofanirwa kuongorora uye kuverengera zvese zvinobvira mamiriro mukati mechikamu chekudzidzisa, izvo zvinotora nguva yakati.
mhedziso
Kubudirira kwekushandisa data ndiyo kiyi yekumisikidza makwikwi mune imwe musika.
Iwe unogona kupatsanura iyo data uchishandisa isina kutariswa muchina kudzidza maalgorithms kuti uongorore zvaunofarira zvevateereri vako kana kuona kuti hutachiona hunopindura sei kune kumwe kurapwa.
Kune akati wandei anoshanda maapplication, uye data masayendisiti, mainjiniya, uye vagadziri vezvivakwa vanogona kukubatsira mukutsanangura zvinangwa zvako uye kugadzira yakasarudzika ML mhinduro kukambani yako.
Leave a Reply