Ungathanda ukuqalisa ukufunda imishini?
Ngidale isifundo esilula nesilula sabaqalayo abaphelele. Ngokuhlangene, sizodlula izinyathelo eziyisisekelo zokuqeqesha imodeli yokufunda yomshini.
Ngenkathi ngichaza izinyathelo zokuqeqesha imodeli ngayinye ngayinye, ngizophinde nginikeze isibonelo esiyisisekelo senkinga yokufunda umshini futhi. Ngakho-ke, uma ungathanda ukulandela, ungalanda lesethi yedatha yesampula kusuka kulokhu isixhumanisi.
Lena idathasethi yesampula nje yokukusiza ukuthi uqalise ngokufunda komshini.
Sinamanani angu-18 abantu bobudala obuhlukene nobulili obuchazwe umculo abawuthandayo. Ngokusebenzisa, izici "zeminyaka" kanye "nobulili" sizozama ukuqagela ukuthi yiluphi uhlobo lomculo abaluthandayo.
Qaphela: 1 kanye no-0 babelwa ubulili njengowesifazane nowesilisa kule dathasethi.
Nokho, uma ungafuni ukulandela isibonelo, nakho kuhle ngokuphelele. Ngizobe ngichaza zonke lezi zinyathelo ngokuningiliziwe. Ngakho-ke, ake sicwilise!
Izinto Zokuqala Okufanele Uzazi
Ngaphambi kokungena ezinyathelweni zokuqeqesha imodeli, ake sicacise amaphuzu athile. Ukufunda ngomshini kuyinto ukuhlakanipha okungekhona okwangempela isiyalo esigxile ekuthuthukiseni ama-algorithms angafunda kudatha.
Ukwenza lokhu, amamodeli okufunda omshini aqeqeshwa kudathasethi efundisa imodeli indlela yokuqagela okulungile noma ngezigaba kudatha entsha, ebingaziwa ngaphambilini.
Ngakho, ayini lawa mamodeli? A imodeli yokufunda imishini ifana neresiphi esetshenziswa ikhompuyutha ukwenza izibikezelo zedatha noma ukukhetha.
Imodeli, njengeresiphi, ilandela isethi yemiyalelo yokuhlola idatha futhi ikhiqize izibikezelo noma izahlulelo ngokusekelwe kumaphethini atholakala kudatha. Uma imodeli iqeqeshwa ngedatha eyengeziwe, yilapho izibikezelo zayo ziba nembe kakhulu.
Hlobo Luni Lwamamodeli Esingawaqeqesha?
Ake sibone ukuthi yimaphi amamodeli wokufunda womshini ayisisekelo.
- Ukuhlehla Komugqa: imodeli ebikezela ukuguquguquka okuqhubekayo kwethagethi kusuka kokuguquguqukayo kokufaka okukodwa noma ngaphezulu.
- I-Neural Networks: inethiwekhi yamanodi axhunyiwe angafunda ukuthola amaphethini ayinkimbinkimbi kudatha.
- Izihlahla Zezinqumo: indlela yokuthatha izinqumo eyakhelwe phezu kochungechunge lwezitatimende zegatsha uma kungenjalo.
- Ukuhlanganisa: isethi yamamodeli aqoqa amaphuzu edatha aqhathanisekayo ngokusekelwe ekufananeni.
- I-Logistic Regression: imodeli yezinkinga zokuhlukanisa kanambambili lapho okuguquguqukayo okuhlosiwe kunamanani amabili angaba khona.
- Izihlahla Zezinqumo: indlela yokuthatha izinqumo eyakhelwe phezu kochungechunge lwezitatimende zegatsha uma kungenjalo.
- Ihlathi Elingahleliwe: imodeli ehlanganisiwe eyakhiwe izihlahla eziningi zokunquma. Zivame ukusetshenziselwa izinhlelo zokusebenza zokuhlela nokuhlehla.
- I-K-Nearest Neighbors: imodeli ebikezela ukuhluka okuqondiwe kusetshenziswa amaphuzu edatha aseduze kuka-k kusethi yokuqeqeshwa.
Ngokuya ngenkinga yethu nedathasethi, sinquma ukuthi iyiphi imodeli yokufunda yomshini elingana nesimo sethu kakhulu. Nokho, sizobuyela kulokhu kamuva. Manje, ake siqale ukuqeqesha imodeli yethu. Ngethemba ukuthi usuvele uyilandile ifayela idathasethi uma ungathanda ukulandela isibonelo sethu.
Futhi, ngincoma ukuthi ube Incwadi kaJupyter efakwe emshinini wangakini futhi uwusebenzisele amaphrojekthi akho okufunda ngomshini.
1: Chaza inkinga
Isigaba sokuqala ku ukuqeqesha umshini wokufunda imodeli ichaza inkinga okufanele ixazululwe. Lokhu kuhlanganisa ukukhetha okuguquguqukayo ofisa ukukubikezela (okwaziwa njengokuguquguquka okuqondiwe) kanye neziguquguqukayo ezizosetshenziswa ukukhiqiza lezo zibikezelo (okwaziwa njengezici noma izibikezelo).
Kufanele futhi unqume ukuthi hlobo luni lwenkinga yokufunda ngomshini ozama ukuyilungisa (ukuhlelwa, ukuhlehla, ukuhlanganisa, njalo njalo) nokuthi yiluphi uhlobo lwedatha ozodinga ukuyiqoqa noma ukuyithola ukuze uqeqeshe imodeli yakho.
Uhlobo lwemodeli oyisebenzisayo luzonqunywa uhlobo lwenkinga yokufunda yomshini ohlose ukuyixazulula. Ukuhlukanisa, ukuhlehla, kanye nokuhlanganisa izigaba ezintathu eziyinhloko izinselele zokufunda komshini. Uma ufuna ukubikezela ukuguquguquka kwesigaba, njengokuthi i-imeyili iwugaxekile noma cha, usebenzisa ukuhlukanisa.
Uma ufisa ukubikezela ukuguquguquka okuqhubekayo, njengentengo yendlu, usebenzisa ukwehla. I-Clustering isetshenziselwa ukuhlanganisa izinto zedatha eziqhathanisekayo ngokusekelwe kokufana kwazo.
Uma sibheka isibonelo sethu; Inselele yethu iwukunquma isitayela somculo esithandwayo somuntu kusukela kubulili neminyaka yakhe. Sizosebenzisa isethi yedatha yabantu abangu-18 kulesi sibonelo kanye nolwazi ngeminyaka yabo, ubulili, nesitayela somculo esiyintandokazi.
2. Lungiselela idatha
Ngemva kokuba ucacise inkinga, uzodinga ukulungiselela idatha yokuqeqesha imodeli. Lokhu kuhlanganisa ukuhlanza nokucubungula idatha. Ngakho-ke, singaqinisekisa ukuthi isesimweni lapho i- i-algorithm yokufunda yomshini angasebenzisa.
Lokhu kungase kuhlanganise imisebenzi efana nokususa amanani angekho, ukuguqula idatha yesigaba ibe idatha yezinombolo, nokukala noma ukwenza idatha ibe evamile ukuze kuqinisekiswe ukuthi zonke izici zisezingeni elifanayo.
Isibonelo, nansi indlela osusa ngayo amanani angekho:
import pandas as pd
# Load the data into a pandas DataFrame
data = pd.read_csv('data.csv')
# Check for missing values
print(data.isnull().sum())
# Drop rows with missing values
data.dropna(inplace=True)
# Check that all missing values have been removed
print(data.isnull().sum())
Inothi elincane: Emgqeni o "import pandas as pd",
singenisa umtapo wezincwadi wePandas futhi siwunikeze igama elithi “pd” ukuze kube lula ukubhekisela emisebenzini yawo nezinto kamuva kukhodi.
I-Pandas iyimojuli eyaziwa kakhulu ye-Python yokukhohlisa nokuhlaziya idatha, ikakhulukazi uma isebenza ngedatha ehlelekile noma yethebula.
Esibonelweni sethu sokunquma izinhlobo zomculo. Sizoqala ngokungenisa idathasethi. Ngiyiqambe ngokuthi i-music.csv, nokho, ungayiqamba ngendlela ofuna ngayo.
Ukuze silungiselele idatha yokuqeqesha imodeli yokufunda yomshini, siyihlukanisa ibe izibaluli (iminyaka yobudala nobulili) nezinjongo (uhlobo lomculo).
Sizophinda futhi sihlukanise idatha ibe amasethi okuqeqesha nokuhlola angu-80:20 ukuze sihlole ukusebenza kwemodeli yethu futhi sigweme ukufakwa ngokweqile.
# Import necessary libraries
import pandas as pd
from sklearn.model_selection import train_test_split
# Load data from CSV file/code>
music_data = pd.read_csv('music.csv')
# Split data into features and target
X = music_data.drop(columns=['genre'])
y = music_data['genre']
# Split data into training and testing sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
3. Khetha imodeli yokufunda yomshini.
Ngemva kokulungisa idatha, kufanele ukhethe imodeli yokufunda ngomshini elungele umsebenzi wakho.
Kunama-algorithms ambalwa ongakhetha kuwo, njengezihlahla zesinqumo, ukuhlehla kwezinto, imishini yokusekela i-vector, amanethiwekhi e-neural, nokunye. I-algorithm oyikhethayo izonqunywa uhlobo lwenkinga ozama ukuyiphendula, uhlobo lwedatha onalo, kanye nezidingo zakho zokusebenza.
Sizosebenzisa isigaba sesihlahla sesinqumo kulesi sibonelo ngoba sisebenza nenkinga yokuhlukanisa (ukubikezela idatha yesigaba).
# Import necessary libraries
from sklearn.tree import DecisionTreeClassifier
Nakhu okubonwayo kokuthi I-Decision Tree Classifier isebenza kanjani:
4. Qeqesha imodeli
Ungaqala ukuqeqesha imodeli uma ukhethe i-algorithm eyamukelekayo yokufunda umshini. Lokhu kuhlanganisa ukusebenzisa idatha ekhiqizwe ngaphambilini ukufundisa i-algorithm yokuthi kwenziwa kanjani ukuqagela kudatha entsha, ebingabonwa ngaphambilini.
I-algorithm izoshintsha amapharamitha ayo angaphakathi phakathi nokuqeqeshwa ukuze kuncishiswe umehluko phakathi kwamanani ayo abikezelwe kanye namanani angempela kudatha yokuqeqeshwa. Inani ledatha elisetshenziselwa ukuqeqeshwa, kanye nemingcele ethile ye-algorithm, konke kungaba nomthelela ekunembeni kwemodeli yomphumela.
Esibonelweni sethu esiqondile, njengoba manje sesinqume indlela, singaqeqesha imodeli yethu ngedatha yokuqeqeshwa.
# Train the decision tree classifier
model = DecisionTreeClassifier()
model.fit(X_train, y_train)
5. Linganisa imodeli
Ngemuva kokuthi imodeli isiqeqeshiwe, kufanele ihlolwe kudatha entsha ukuze kuqinisekiswe ukuthi inembile futhi ithembekile. Lokhu kuhlanganisa ukuhlola imodeli ngedatha engazange isetshenziswe ngesikhathi sokuqeqeshwa nokuqhathanisa amanani ayo aqageliwe namanani angempela kudatha yokuhlola.
Lokhu kubuyekezwa kungasiza ekuboneni noma yimaphi amaphutha emodeli, njengokufakela ngokweqile noma ukufakwa ngaphansi, futhi kungaholela kunoma yikuphi ukulungisa okungase kudingeke.
Sisebenzisa idatha yokuhlola, sizohlola ukunemba kwemodeli yethu.
# Import necessary libraries
from sklearn.metrics import accuracy_score
# Predict the music genre for the test data
predictions = model.predict(X_test)
# Evaluate the model's accuracy
accuracy = accuracy_score(y_test, predictions)
print("Accuracy: ", accuracy)
Umphumela wokunemba awumbi kangako okwamanje. 🙂 Ukuze uthuthukise isikolo sakho sokunemba, ungakwazi njalo ukuhlanza idatha kakhulu noma uzame amamodeli okufunda ngomshini ahlukene ukuze ubone ukuthi iyiphi enikeza amaphuzu aphezulu.
6. Lungisa kahle imodeli
Uma ukusebenza kahle kwemodeli kunganele ngokwanele, ungayishuna kahle ngokushintsha amapharamitha ahlukahlukene we-algorithm noma ngokuzama ama-algorithms amasha ngokuphelele.
Le nqubo ingase ihlanganise ukuhlola ezinye izilinganiso zokufunda, ukulungisa izilungiselelo zokujwayela, noma ukushintsha inombolo noma usayizi wezendlalelo ezifihliwe kunethiwekhi ye-neural.
7. Sebenzisa imodeli
Uma usujabule ngokusebenza kwemodeli, ungaqala ukuyisebenzisa ukuze ukhiqize ukuqagela kudatha entsha.
Lokhu kungase kuhlanganise ukuphakela idatha entsha kumodeli nokusebenzisa amapharamitha afundiwe emodeli ukuze kwenziwe ukuqagela kuleyo datha, noma ukuhlanganisa imodeli ibe uhlelo lokusebenza olubanzi noma isistimu.
Singasebenzisa imodeli yethu ukuze sikhiqize izibikezelo kudatha entsha ngemva kokuthi sijabule ngokunemba kwayo. Ungazama amanani ahlukene obulili neminyaka.
# Test the model with new data
new_data = [[25, 1], [30, 0]]
predictions = model.predict(new_data)
print("Predictions: ", predictions)
Qedani
Sesiqedile ukuqeqesha imodeli yethu yokuqala yokufunda umshini.
Ngethemba ukuthi ukuthole kuwusizo. Manje usungazama ukusebenzisa amamodeli okufunda emishini ahlukene njenge-Linear Regression noma i-Random Forest.
Kunamasethi edatha amaningi nezinselele kuwo Igagasi uma ungathanda ukuthuthukisa ukubhala kwakho ikhodi nokuqonda kokufunda komshini.
shiya impendulo