Model selection و tuning في scikit-learn

TRAIN          → CV/TUNING           → FINAL TEST
(~70%)         (validation, ~15%)      (held-out, ~15%)

from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV, train_test_split
from sklearn.metrics import accuracy_score
import numpy as np

# 1. بيانات
X = np.random.random((200, 5))
y = np.random.randint(0, 2, 200)

# 2. تقسيم
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

# 3. تعريف فضاء البحث
param_grid = {
    'n_estimators': [50, 100, 200],
    'max_depth': [3, 5, 10, None],
    'min_samples_split': [2, 5, 10]
}

# 4. GridSearchCV
grid = GridSearchCV(
    estimator=RandomForestClassifier(random_state=42),
    param_grid=param_grid,
    cv=5,                  # 5-fold CV لكل توليفة
    scoring='accuracy',
    n_jobs=-1              # كل الـ CPU cores
)

grid.fit(X_train, y_train)

# 5. أفضل توليفة
print(f"أفضل توليفة: {grid.best_params_}")
print(f"أفضل CV score: {grid.best_score_:.3f}")

# 6. تقييم نهائي على test
best_model = grid.best_estimator_
y_pred = best_model.predict(X_test)
print(f"Test accuracy: {accuracy_score(y_test, y_pred):.3f}")

from sklearn.model_selection import RandomizedSearchCV
from scipy.stats import randint

param_distributions = {
    'n_estimators': randint(50, 300),
    'max_depth': randint(3, 20),
    'min_samples_split': randint(2, 20)
}

random_search = RandomizedSearchCV(
    estimator=RandomForestClassifier(random_state=42),
    param_distributions=param_distributions,
    n_iter=20,             # 20 توليفة عشوائية فقط
    cv=5,
    scoring='accuracy',
    n_jobs=-1,
    random_state=42
)

random_search.fit(X_train, y_train)
print(f"أفضل توليفة: {random_search.best_params_}")
print(f"أفضل CV score: {random_search.best_score_:.3f}")

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import GridSearchCV

pipe = Pipeline([
    ('scaler', StandardScaler()),
    ('model', LogisticRegression(max_iter=1000))
])

# الـ hyperparameters تستخدم 'model__' prefix
param_grid = {
    'model__C': [0.1, 1.0, 10.0],
    'model__penalty': ['l1', 'l2']
}

grid = GridSearchCV(pipe, param_grid, cv=5, scoring='accuracy')
grid.fit(X_train, y_train)

Tuning iteration لكل توليفة:
  Fold 1: train → val → score₁
  Fold 2: train → val → score₂
  Fold 3: train → val → score₃
  Fold 4: train → val → score₄
  Fold 5: train → val → score₅
  → متوسط الـ 5 scores → score_cv

اختيار أفضل توليفة: أعلى score_cv.

# ❌ خطأ
for C in [0.1, 1.0, 10.0]:
    model = LogisticRegression(C=C)
    model.fit(X_train, y_train)
    score = model.score(X_test, y_test)  # ← leakage!
    # "أفضل" C هو الذي يحقّق أعلى score على test

# 1. CV/Tuning على train فقط
grid = GridSearchCV(model, param_grid, cv=5)
grid.fit(X_train, y_train)

# 2. تقييم نهائي على test — مرة واحدة
best_model = grid.best_estimator_
final_score = best_model.score(X_test, y_test)

from sklearn.pipeline import Pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import GridSearchCV, train_test_split, cross_val_score
from sklearn.metrics import accuracy_score, classification_report
import numpy as np

# 1. بيانات
X, y = ...  # بياناتك
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

# 2. Pipeline
pipe = Pipeline([
    ('scaler', StandardScaler()),
    ('model', RandomForestClassifier(random_state=42))
])

# 3. فضاء البحث
param_grid = {
    'model__n_estimators': [100, 200],
    'model__max_depth': [5, 10, None]
}

# 4. Tuning
grid = GridSearchCV(pipe, param_grid, cv=5, scoring='accuracy', n_jobs=-1)
grid.fit(X_train, y_train)

# 5. أفضل توليفة
print(f"Best: {grid.best_params_}")
print(f"Best CV score: {grid.best_score_:.3f}")

# 6. تقييم نهائي على test
best_model = grid.best_estimator_
y_pred = best_model.predict(X_test)

print(f"\nFinal test accuracy: {accuracy_score(y_test, y_pred):.3f}")
print(classification_report(y_test, y_pred))