Scikit-learn Tutorial: Machine Learning in Python from Scratch (2026)
Scikit-learn is the most widely used ML library in Python. It gives you clean, consistent APIs for dozens of algorithms, plus preprocessing, model selection, and evaluation tools. This guide covers everything from raw data to a deployed model.
The fit / predict / transform API
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler
model = LogisticRegression()
model.fit(X_train, y_train) # learn from training data
preds = model.predict(X_test) # make predictions
proba = model.predict_proba(X_test) # class probabilities
scaler = StandardScaler()
scaler.fit(X_train) # compute mean/std from train set only
X_train_s = scaler.transform(X_train)
X_test_s = scaler.transform(X_test) # reuse train stats on testEnd-to-End Example
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import classification_report
data = load_breast_cancer(as_frame=True)
X, y = data.data, data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)
rf = RandomForestClassifier(n_estimators=100, random_state=42)
rf.fit(X_train, y_train)
print('RF Acc:', rf.score(X_test, y_test))
print(classification_report(y_test, rf.predict(X_test), target_names=data.target_names))Pipelines: Prevent Data Leakage
from sklearn.pipeline import Pipeline
from sklearn.svm import SVC
# Correct -- Pipeline fits scaler only on training data:
pipe = Pipeline([
('scaler', StandardScaler()),
('svm', SVC(kernel='rbf', probability=True))
])
pipe.fit(X_train, y_train)
print('Pipeline Acc:', pipe.score(X_test, y_test))Cross-Validation
from sklearn.model_selection import cross_val_score, StratifiedKFold
cv = StratifiedKFold(n_splits=5, shuffle=True, random_state=42)
scores = cross_val_score(pipe, X, y, cv=cv, scoring='accuracy')
print(f'CV Accuracy: {scores.mean():.4f} +/- {scores.std():.4f}')Preprocessing Toolkit
from sklearn.compose import ColumnTransformer
from sklearn.impute import SimpleImputer
from sklearn.preprocessing import OneHotEncoder
preprocessor = ColumnTransformer([
('num', Pipeline([('imp', SimpleImputer(strategy='median')), ('sc', StandardScaler())]), ['age','income']),
('cat', Pipeline([('imp', SimpleImputer(strategy='most_frequent')), ('enc', OneHotEncoder(handle_unknown='ignore'))]), ['city','gender']),
])Hyperparameter Tuning
from sklearn.model_selection import GridSearchCV
param_grid = {'svm__C':[0.1,1,10], 'svm__gamma':['scale','auto',0.01,0.1]}
gs = GridSearchCV(pipe, param_grid, cv=5, n_jobs=-1)
gs.fit(X_train, y_train)
print('Best:', gs.best_params_)10 Algorithms to Know
LogisticRegression for fast interpretable classification. Ridge and Lasso for regularised regression. SVC for powerful medium-size datasets. DecisionTreeClassifier for interpretability. RandomForestClassifier as an excellent default. GradientBoostingClassifier for powerful boosting. KNeighborsClassifier for simple baselines. GaussianNB for fast text. KMeans for clustering. PCA for dimensionality reduction.
Conclusion
Master fit/predict/transform, use Pipelines to prevent data leakage, cross-validate for reliable evaluation, and GridSearchCV for tuning. These four concepts let you apply any of scikit-learn’s 50+ estimators confidently.
