M'ndandanda wazopezekamo[Bisani][Show]
Ngati ndinu wopanga mapulogalamu a Python kapena mukufuna zida zamphamvu zomwe mungagwiritse ntchito kuyambitsa kuphunzira pamakina pamakina opangira, Scikit-learn ndi laibulale yomwe muyenera kuyang'ana.
Scikit-lern ndi yolembedwa bwino komanso yosavuta kugwiritsa ntchito, kaya mwangoyamba kumene kuphunzira pamakina, mukufuna kudzuka ndikuthamanga mwachangu, kapena mukufuna kugwiritsa ntchito chida chofufuzira chaposachedwa kwambiri cha ML.
Zimakulolani kuti mupange chitsanzo cha deta yolosera m'mizere yochepa chabe ya code ndikugwiritsira ntchito chitsanzocho kuti chigwirizane ndi deta yanu monga laibulale yapamwamba. Ndi yosinthika ndipo imagwira ntchito bwino ndi ena Python library monga Matplotlib ya charting, NumPy for array vectorization, and pandas for data visualization.
Mu bukhuli, mupeza zonse zomwe zili, momwe mungagwiritsire ntchito, komanso zabwino ndi zoyipa zake.
Kodi Scikit-phunzirani?
Scikit-learn (yomwe imadziwikanso kuti sklearn) imapereka mitundu yosiyanasiyana ya ziwerengero ndi kuphunzira pamakina. Mosiyana ndi ma modules ambiri, sklearn imapangidwa ku Python m'malo mwa C. Ngakhale kuti ikupangidwa ku Python, mphamvu ya sklearn imatchulidwa kuti imagwiritsa ntchito NumPy pakuchita bwino kwambiri kwa algebra ndi machitidwe osiyanasiyana.
Scikit-Learn idapangidwa ngati gawo la polojekiti ya Google ya Summer of Code ndipo kuyambira pamenepo yapangitsa miyoyo ya mamiliyoni a asayansi a Python-centric padziko lonse kukhala osavuta. Gawo ili la mndandanda likuyang'ana kwambiri pakuwonetsa laibulale ndikuwunikira chinthu chimodzi - kusintha kwa dataset, komwe ndi gawo lofunikira komanso lofunikira lomwe muyenera kuchita musanapange chitsanzo cholosera.
Laibulaleyi idakhazikitsidwa ndi SciPy (Scientific Python), yomwe iyenera kukhazikitsidwa musanagwiritse ntchito scikit-lern. Mundawuwu uli ndi zinthu izi:
- NumPy: Phukusi la Python la n-dimensional array
- SciPy: Ndi phukusi lofunikira pamakompyuta asayansi
- Pandas: Mapangidwe a data ndi kusanthula
- Matplotlib: Ndi laibulale yamphamvu yopangira 2D/3D
- Sympy: Masamu ophiphiritsa
- IPython: Kuwongolera kolumikizirana
Kugwiritsa ntchito laibulale ya Scikit-lern
Scikit-learn ndi phukusi lotseguka la Python lomwe lili ndi kusanthula kwatsatanetsatane kwa data komanso mawonekedwe amigodi. Zimabwera ndi kuchuluka kwa ma aligorivimu opangidwa kuti akuthandizeni kuti mupindule kwambiri ndi mapulojekiti anu a sayansi ya data. Laibulale yophunzirira ya Scikit imagwiritsidwa ntchito motere.
1. Kuponderezana
Kusanthula kwa regression ndi njira yowerengera ndikuwunikira kugwirizana pakati pa mitundu iwiri kapena kuposerapo. Njira yomwe imagwiritsidwa ntchito poyesa kuyambiranso imathandizira kudziwa kuti ndi zinthu ziti zomwe zili zoyenera, zomwe zinganyalanyazidwe, komanso momwe zimalumikizirana. Njira zochepetsera, mwachitsanzo, zitha kugwiritsidwa ntchito kuti mumvetsetse bwino momwe mitengo yamitengo imayendera.
Regression algorithms ndi:
- Kugonjetsa kwapakati
- Ridge Regression
- Kuponderezedwa kwa Lasso
- Kubwerera kwa Mtengo Wachigamulo
- Random Forest
- Makina Othandizira Vector (SVM)
2. Gulu
Njira Yoyang'anira ndi Njira Yophunzirira Yoyang'aniridwa yomwe imagwiritsa ntchito deta yophunzitsira kuti izindikire gulu lazowunikira zatsopano. Algorithm mu Gulu imaphunzira kuchokera ku zomwe zaperekedwa pepala kapena kuwunika ndikuyika zowunikira zina kukhala gulu limodzi mwamagulu ambiri. Atha, mwachitsanzo, kugwiritsidwa ntchito kuyika mauthenga a imelo ngati sipamu kapena ayi.
Classification algorithms ndi awa:
- Kukonzanso Zinthu
- K-Oyandikana nawo Apafupi
- Makina Othandizira Vector
- Mtengo Wosankha
- Random Forest
3. Kusonkhanitsa
Ma clustering ma algorithms mu Scikit-learn amagwiritsidwa ntchito kusanja deta yokhala ndi zinthu zofanana kukhala ma seti. Clustering ndi njira yoyika zinthu m'magulu kuti zomwe zili m'gulu limodzi zifanane ndi zomwe zili m'magulu ena. Zambiri zamakasitomala, mwachitsanzo, zitha kulekanitsidwa kutengera komwe ali.
Ma Clustering algorithms ndi awa:
- DB-SCAN
- K-njira
- Mini-Batch K-Njira
- Kuphatikiza kwa Spectral
4. Kusankha Chitsanzo
Ma algorithms osankha zitsanzo amapereka njira zofananira, zotsimikizira, ndikusankha magawo ndi zitsanzo zomwe zingagwiritsidwe ntchito poyambira sayansi ya data. Kutengera deta, kusankha kwachitsanzo ndi vuto la kusankha chitsanzo chowerengera kuchokera ku gulu la anthu ofuna kusankha. Muzochitika zofunikira kwambiri, kusonkhanitsa deta komwe kunalipo kale kumaganiziridwa. Komabe, ntchitoyi ingaphatikizeponso kupanga zoyesera kuti deta yomwe yapezedwa igwirizane bwino ndi vuto la kusankha chitsanzo.
Ma module osankhidwa omwe angapangitse kulondola posintha magawo ndi awa:
- Kutsimikizira-mtanda
- Kusaka kwa Gridi
- Miyala
5. Kuchepetsa Dimensionality
Kusamutsidwa kwa deta kuchokera ku malo apamwamba kupita kumalo otsika kwambiri kotero kuti mawonekedwe otsika kwambiri asunge mbali zina zofunika za deta yoyambirira, makamaka pafupi ndi chikhalidwe chake, amadziwika kuti kuchepetsa dimensionality. Chiwerengero cha zosintha zosasinthika kuti ziwunikidwe zimachepetsedwa pamene kukula kumachepetsedwa. Zidziwitso zakunja, mwachitsanzo, sizingaganizidwe kuti zitha kupititsa patsogolo luso la zowonera.
Dimensionality Reduction algorithm ili ndi izi:
- Kusankha kwamachitidwe
- Kuwunika Kwakukulu Kwambiri (PCA)
Kukhazikitsa Scikit-learn
NumPy, SciPy, Matplotlib, IPython, Sympy, ndi Pandas akuyenera kukhazikitsidwa musanagwiritse ntchito Scikit-learn. Tiyeni tiyike pogwiritsa ntchito pip kuchokera ku console (imagwira ntchito pa Windows yokha).
Tiyeni tiyike Scikit-phunzirani tsopano popeza tayika malaibulale ofunikira.
Mawonekedwe
Scikit-lern, yomwe nthawi zina imadziwika kuti sklearn, ndi chida cha Python chogwiritsa ntchito makina ophunzirira makina ndi mawerengero a mawerengero. Titha kuzigwiritsa ntchito popanga mitundu ingapo yophunzirira makina osinthira, kugawika, ndi kusanja, komanso zida zowerengera zowunikira mitundu iyi. Zimaphatikizanso kuchepetsa kukula, kusankha mawonekedwe, kutulutsa mawonekedwe, njira zophatikizira, ndi ma dataset omangidwa. Tidzafufuza lililonse la makhalidwe amenewa limodzi ndi limodzi.
1. Kuitanitsa Ma Dataset
Kuphunzira kwa Scikit kumaphatikizapo ma dataset angapo omwe adamangidwa kale, monga dataset ya iris, dataset yamitengo yanyumba, titanic dataset, ndi zina zotero. Ubwino waukulu wamagulu awa ndikuti ndi osavuta kumva ndipo atha kugwiritsidwa ntchito kupanga mitundu ya ML nthawi yomweyo. Ma dataset awa ndi oyenera kwa oyambira. Momwemonso, mutha kugwiritsa ntchito sklearn kuitanitsa ma dataset owonjezera. Momwemonso, mutha kuyigwiritsa ntchito kuyitanitsa ma dataset owonjezera.
2. Kugawanitsa Dataset kwa Maphunziro ndi Kuyesa
Sklearn idaphatikizanso kuthekera kogawa deta mu magawo ophunzitsira ndi kuyesa. Kugawanitsa deta ndikofunikira kuti muwunikire mopanda tsankho la magwiridwe antchito. Titha kufotokoza kuchuluka kwa deta yathu yomwe iyenera kuphatikizidwa mu masitima apamtunda ndi ma dataset oyeserera. Tidagawanitsa deta pogwiritsa ntchito kuyesa kwa masitima apamtunda kotero kuti sitimayo imakhala ndi 80% ya data ndipo yoyeserera imakhala ndi 20%. Dataset ikhoza kugawidwa motere:
3. Linear Regression
Linear Regression ndi njira yoyang'aniridwa yophunzirira pamakina. Imagwira ntchito yobwezeretsa. Kutengera zosinthika zodziyimira pawokha, kutsika kumawonetsa mtengo wolosera zolinga. Nthawi zambiri amagwiritsidwa ntchito kudziwa kugwirizana pakati pa zosintha ndi kulosera. Mitundu yosiyanasiyana yobwereranso imasiyana malinga ndi mtundu wa kulumikizana komwe amawunika pakati pa zodalira ndi zodziyimira pawokha, komanso kuchuluka kwa zosintha zodziyimira pawokha zomwe zimagwiritsidwa ntchito. Titha kungopanga mtundu wa Linear Regression pogwiritsa ntchito sklearn motere:
4. Kubwerera kwa Logistic
Njira yodziwika bwino yogawa magawo ndikusintha kwazinthu. Zili m'banja lomwelo monga polynomial ndi linear regression ndipo ndi m'gulu la mzere wamagulu. Zomwe zapezeka pakubweza kwa logistic ndizosavuta kumvetsetsa komanso zimawerengera mwachangu. Momwemonso kubwezeredwa kwa mzere, logistic regression ndi njira yoyang'aniridwa yobwereranso. Zosintha zotulutsa ndizosiyana, ndiye kusiyana kokhako. Ikhoza kudziwa ngati wodwala ali ndi matenda a mtima kapena ayi.
Nkhani zosiyanasiyana zamagulu, monga kuzindikira sipamu, zitha kuthetsedwa pogwiritsa ntchito njira yobwerera. Kulosera kwa matenda a shuga, kudziwa ngati wogula angagule chinthu china kapena kusinthana ndi mnzake, kudziwa ngati wogwiritsa ntchito adina ulalo winawake wamalonda, ndipo zina zambiri ndi zitsanzo zochepa chabe.
5. Mtengo Wosankha
Njira yamphamvu kwambiri komanso yogwiritsidwa ntchito kwambiri m'magulu ndi kulosera ndi mtengo wazisankho. Mtengo wosankha ndi mtengo womwe umawoneka ngati tchati, ndi mfundo iliyonse yamkati ikuyimira mayeso pamalingaliro, nthambi iliyonse imayimira mapeto a mayeso, ndi tsamba lililonse (terminal node) yokhala ndi lebulo la kalasi.
Ngati mitundu yodalira ilibe mgwirizano wa mzere ndi zosinthika zodziyimira pawokha, mwachitsanzo, ngati kusintha kwa mzere sikutulutsa zolondola, mitengo yachisankho imakhala yopindulitsa. Chinthu cha DecisionTreeRegression() chitha kugwiritsidwa ntchito mofananamo kugwiritsa ntchito mtengo wachigamulo kuti ubwererenso.
6. Nkhalango Yachisawawa
Nkhalango mwachisawawa ndi makina kuphunzira njira yothetsera mavuto obwerera m'mbuyo ndi magulu. Zimagwiritsa ntchito kuphunzira pamodzi, yomwe ndi njira yomwe imaphatikiza magulu angapo kuti athetse mavuto ovuta. Njira yachisawawa ya nkhalango imapangidwa ndi mitengo yambiri yosankha. Itha kugwiritsidwa ntchito kugawa zopempha za ngongole m'magulu, kuzindikira zachinyengo, ndi kuyembekezera kubuka kwa matenda.
7. Chisokonezo Matrix
A confusion matrix ndi tebulo lomwe limagwiritsidwa ntchito pofotokozera machitidwe a magulu. Mawu anayi otsatirawa amagwiritsidwa ntchito pofufuza chisokonezo:
- Zoona Zowona: Zikutanthauza kuti chitsanzocho chinapereka zotsatira zabwino ndipo zinali zolondola.
- Zolakwika Zowona: Zikutanthauza kuti chitsanzocho chinawonetsa zotsatira zoipa ndipo zinali zolondola.
- Zonama Zonama: Zikutanthauza kuti chitsanzocho chinkayembekezera zotsatira zabwino koma zinali zoipa.
- Zolakwika Zonama: Zikutanthauza kuti chitsanzocho chinkayembekezera zotsatira zoipa, pamene zotsatira zake zinali zabwino.
Kukhazikitsa kosokoneza matrix:
ubwino
- Ndi yosavuta kugwiritsa ntchito.
- Phukusi la Scikit-lern ndi losinthika kwambiri komanso lothandiza, likutumikira zolinga zenizeni zenizeni monga kuneneratu za khalidwe la ogula, chitukuko cha neuroimage, ndi zina zotero.
- Ogwiritsa ntchito omwe akufuna kulumikiza ma aligorivimu ndi nsanja zawo apeza zolemba zatsatanetsatane za API patsamba la Scikit-lern.
- Olemba ambiri, ogwira nawo ntchito, komanso gulu lalikulu lapaintaneti padziko lonse lapansi ndikuthandizira Scikit-phunzirani zatsopano.
kuipa
- Si njira yabwino yophunzirira mozama.
Kutsiliza
Scikit-lern ndi phukusi lofunikira kuti wasayansi aliyense azidziwa bwino komanso kudziwa zambiri. Bukuli likuyenera kukuthandizani pakusokoneza deta pogwiritsa ntchito sklearn. Pali zina zambiri za Scikit-phunzirani zomwe mudzazipeza mukamapita patsogolo paulendo wanu wa sayansi ya data. Gawani maganizo anu mu ndemanga.
Siyani Mumakonda