from sklearn.linear_model import LogisticRegression

# 创建模型
model = LogisticRegression(
    penalty='l2',           # 正则化类型: 'l1', 'l2', 'elasticnet', None
    C=1.0,                  # 正则化强度的倒数（越小正则化越强）
    solver='lbfgs',         # 优化算法
    max_iter=1000,          # 最大迭代次数
    multi_class='auto',     # 多分类策略
    class_weight=None,      # 类别权重（处理不平衡数据）
    random_state=42
)

# 训练
model.fit(X_train, y_train)

# 预测
y_pred = model.predict(X_test)          # 类别预测
y_prob = model.predict_proba(X_test)    # 概率预测

# 评估
accuracy = model.score(X_test, y_test)

from sklearn.metrics import classification_report

print(classification_report(y_test, y_pred, target_names=['负类', '正类']))

"""
Day 108 实战：使用逻辑回归进行乳腺癌诊断分类
数据集：sklearn 内置乳腺癌数据集
"""
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import (
    confusion_matrix, classification_report,
    roc_curve, auc, accuracy_score
)

# ==================== 1. 加载数据 ====================
data = load_breast_cancer()
X, y = data.data, data.target
print(f"特征数: {X.shape[1]}, 样本数: {X.shape[0]}")
print(f"类别分布: 恶性={sum(y==0)}, 良性={sum(y==1)}")

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

# ==================== 3. 特征标准化 ====================
# 逻辑回归对特征尺度敏感，必须标准化
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

# ==================== 4. 训练模型 ====================
model = LogisticRegression(
    penalty='l2',
    C=1.0,
    solver='lbfgs',
    max_iter=1000,
    random_state=42
)
model.fit(X_train_scaled, y_train)

# ==================== 5. 预测与评估 ====================
y_pred = model.predict(X_test_scaled)
y_prob = model.predict_proba(X_test_scaled)[:, 1]  # 正类概率

print(f"\n准确率: {accuracy_score(y_test, y_pred):.4f}")
print(f"\n分类报告:\n{classification_report(y_test, y_pred, target_names=['恶性', '良性'])}")

# ==================== 6. 混淆矩阵 ====================
cm = confusion_matrix(y_test, y_pred)
print(f"混淆矩阵:\n{cm}")

# ==================== 7. ROC 曲线 ====================
fpr, tpr, thresholds = roc_curve(y_test, y_prob)
roc_auc = auc(fpr, tpr)
print(f"\nAUC: {roc_auc:.4f}")

# ==================== 8. 可视化 ====================
fig, axes = plt.subplots(1, 2, figsize=(14, 5))

# 混淆矩阵热力图
im = axes[0].imshow(cm, interpolation='nearest', cmap=plt.cm.Blues)
axes[0].set_title('混淆矩阵')
axes[0].set_xlabel('预测标签')
axes[0].set_ylabel('真实标签')
tick_marks = np.arange(2)
axes[0].set_xticks(tick_marks)
axes[0].set_xticklabels(['恶性', '良性'])
axes[0].set_yticks(tick_marks)
axes[0].set_yticklabels(['恶性', '良性'])
for i in range(2):
    for j in range(2):
        axes[0].text(j, i, format(cm[i, j], 'd'),
                     ha="center", va="center",
                     color="white" if cm[i, j] > cm.max()/2 else "black")
plt.colorbar(im, ax=axes[0])

# ROC 曲线
axes[1].plot(fpr, tpr, color='darkorange', lw=2, label=f'ROC curve (AUC = {roc_auc:.4f})')
axes[1].plot([0, 1], [0, 1], color='navy', lw=2, linestyle='--', label='随机模型')
axes[1].set_xlim([0.0, 1.0])
axes[1].set_ylim([0.0, 1.05])
axes[1].set_xlabel('假正例率 (FPR)')
axes[1].set_ylabel('真正例率 (TPR)')
axes[1].set_title('ROC 曲线')
axes[1].legend(loc="lower right")

plt.tight_layout()
plt.savefig('day-108-evaluation.png', dpi=150, bbox_inches='tight')
print("\n✅ 评估图表已保存")

# ==================== 9. 查看特征重要性 ====================
feature_importance = np.abs(model.coef_[0])
top_indices = np.argsort(feature_importance)[::-1][:10]
print("\nTop 10 重要特征:")
for i, idx in enumerate(top_indices):
    print(f"  {i+1}. {data.feature_names[idx]}: {feature_importance[idx]:.4f}")

"""
Day 108 进阶：正则化强度对模型的影响 + 交叉验证
"""
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler

# 加载并预处理数据
data = load_breast_cancer()
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
)
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

# ==================== 1. 不同 C 值的对比 ====================
C_values = [0.001, 0.01, 0.1, 0.5, 1.0, 5.0, 10.0, 50.0, 100.0]
train_scores = []
test_scores = []
cv_scores = []

for C in C_values:
    model = LogisticRegression(C=C, max_iter=5000, random_state=42)
    model.fit(X_train_scaled, y_train)
    
    train_scores.append(model.score(X_train_scaled, y_train))
    test_scores.append(model.score(X_test_scaled, y_test))
    
    # 5 折交叉验证
    cv = cross_val_score(model, X_train_scaled, y_train, cv=5)
    cv_scores.append(cv.mean())

# ==================== 2. 可视化 ====================
plt.figure(figsize=(10, 6))
plt.plot(C_values, train_scores, 'o-', label='训练集准确率', color='blue')
plt.plot(C_values, test_scores, 's-', label='测试集准确率', color='red')
plt.plot(C_values, cv_scores, '^-', label='5折交叉验证', color='green')
plt.xscale('log')
plt.xlabel('C (正则化强度倒数)')
plt.ylabel('准确率')
plt.title('正则化强度 vs 模型性能')
plt.legend()
plt.grid(True, alpha=0.3)
plt.savefig('day-108-regularization.png', dpi=150, bbox_inches='tight')
print("✅ 正则化对比图已保存")

# ==================== 3. 分析结果 ====================
print("\n📊 各 C 值性能对比:")
print(f"{'C值':>10} {'训练集':>10} {'测试集':>10} {'CV均值':>10}")
print("-" * 45)
for C, tr, te, cv in zip(C_values, train_scores, test_scores, cv_scores):
    print(f"{C:>10.3f} {tr:>10.4f} {te:>10.4f} {cv:>10.4f}")

# ==================== 4. 最佳模型分析 ====================
best_idx = np.argmax(cv_scores)
best_C = C_values[best_idx]
print(f"\n🏆 最佳 C 值: {best_C} (CV 准确率: {cv_scores[best_idx]:.4f})")

# 过拟合判断
if train_scores[best_idx] - test_scores[best_idx] > 0.05:
    print("⚠️  警告：训练集与测试集差距较大，可能存在过拟合")
else:
    print("✅ 模型泛化能力良好")

"""
Day 108 实战：逻辑回归多分类 — 鸢尾花数据集
对比 OvR 和 Softmax 两种策略
"""
import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.linear_model import LogisticRegression
from sklearn.preprocessing import StandardScaler
from sklearn.metrics import confusion_matrix, classification_report
from matplotlib.colors import ListedColormap

# ==================== 1. 加载数据 ====================
iris = load_iris()
X, y = iris.data, iris.target
feature_names = iris.feature_names
target_names = iris.target_names

print(f"数据集: {X.shape[0]} 样本, {X.shape[1]} 特征")
print(f"类别: {list(target_names)}")

# 只用前两个特征方便可视化
X_2d = X[:, :2]
X_train, X_test, y_train, y_test = train_test_split(
    X_2d, y, test_size=0.2, random_state=42, stratify=y
)

scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

# ==================== 2. OvR 策略 ====================
print("\n" + "="*50)
print("📌 策略 1: One-vs-Rest (OvR)")
print("="*50)

ovr_model = LogisticRegression(
    multi_class='ovr',
    solver='lbfgs',
    max_iter=1000,
    random_state=42
)
ovr_model.fit(X_train_scaled, y_train)
y_pred_ovr = ovr_model.predict(X_test_scaled)

print(f"准确率: {ovr_model.score(X_test_scaled, y_test):.4f}")
print(f"\n分类报告:\n{classification_report(y_test, y_pred_ovr, target_names=target_names)}")

# ==================== 3. Softmax 策略 ====================
print("\n" + "="*50)
print("📌 策略 2: Softmax (Multinomial)")
print("="*50)

softmax_model = LogisticRegression(
    multi_class='multinomial',
    solver='lbfgs',
    max_iter=1000,
    random_state=42
)
softmax_model.fit(X_train_scaled, y_train)
y_pred_softmax = softmax_model.predict(X_test_scaled)

print(f"准确率: {softmax_model.score(X_test_scaled, y_test):.4f}")
print(f"\n分类报告:\n{classification_report(y_test, y_pred_softmax, target_names=target_names)}")

# ==================== 4. 决策边界可视化 ====================
fig, axes = plt.subplots(1, 2, figsize=(14, 6))
cmap_light = ListedColormap(['#FFAAAA', '#AAFFAA', '#AAAAFF'])
cmap_bold = ListedColormap(['#FF0000', '#00FF00', '#0000FF'])

for ax, model, title in zip(axes, [ovr_model, softmax_model], ['OvR', 'Softmax']):
    # 创建网格
    h = 0.02
    x_min, x_max = X_train_scaled[:, 0].min() - 1, X_train_scaled[:, 0].max() + 1
    y_min, y_max = X_train_scaled[:, 1].min() - 1, X_train_scaled[:, 1].max() + 1
    xx, yy = np.meshgrid(np.arange(x_min, x_max, h), np.arange(y_min, y_max, h))
    
    # 预测网格点
    Z = model.predict(np.c_[xx.ravel(), yy.ravel()])
    Z = Z.reshape(xx.shape)
    
    ax.contourf(xx, yy, Z, alpha=0.3, cmap=cmap_light)
    scatter = ax.scatter(X_train_scaled[:, 0], X_train_scaled[:, 1],
                        c=y_train, cmap=cmap_bold, edgecolors='k', s=50)
    ax.set_xlabel(feature_names[0])
    ax.set_ylabel(feature_names[1])
    ax.set_title(f'{title} 决策边界')
    
# 添加图例
legend_elements = [plt.Line2D([0], [0], marker='o', color='w', 
                  markerfacecolor=c, markersize=10, label=n)
                  for c, n in zip(['red', 'green', 'blue'], target_names)]
fig.legend(handles=legend_elements, loc='upper center', ncol=3)

plt.tight_layout()
plt.savefig('day-108-multiclass.png', dpi=150, bbox_inches='tight')
print("\n✅ 多分类对比图已保存")

# ==================== 5. 概率输出分析 ====================
print("\n📊 概率输出示例 (Softmax):")
proba = softmax_model.predict_proba(X_test_scaled[:5])
for i in range(5):
    print(f"  样本{i+1}: {dict(zip(target_names, proba[i].round(4)))} → 预测: {target_names[y_pred_softmax[i]]}")

# ❌ 错误：直接训练
model.fit(X_train, y_train)  # 特征尺度差异大时，收敛慢甚至不收敛

# ✅ 正确：先标准化
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
model.fit(X_train_scaled, y_train)

# ❌ 错误：99% 正样本，全预测正类就有 99% 准确率
accuracy = model.score(X_test, y_test)  # 看起来很高，实际没用

# ✅ 正确：用 F1-Score 或 AUC
from sklearn.metrics import f1_score, roc_auc_score
f1 = f1_score(y_test, y_pred)
auc = roc_auc_score(y_test, y_prob)

# 或者用 class_weight 处理不平衡
model = LogisticRegression(class_weight='balanced')

# ❌ 错误：在训练集上评估，结果虚高
train_accuracy = model.score(X_train, y_train)  # 通常很高，但没意义

# ✅ 正确：用测试集或交叉验证
test_accuracy = model.score(X_test, y_test)
cv_scores = cross_val_score(model, X_train, y_train, cv=5)

# ❌ 警告：ConvergenceWarning
model = LogisticRegression(max_iter=100)  # 默认值，可能不够

# ✅ 解决：加大迭代次数
model = LogisticRegression(max_iter=1000)
# 或者更好的做法：先标准化数据