import pandas as pd
import numpy as np
import yfinance as yf
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error
from xgboost import XGBRegressor
import matplotlib.pyplot as plt
import seaborn as sns
import warnings

warnings.filterwarnings('ignore')
# 设置可视化风格
sns.set_style("whitegrid")
plt.rcParams['font.sans-serif'] = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号

# 创建一个模拟数据集，因为 yfinance 在某些环境下可能不稳定
# 真实场景下，应使用下面的 yfinance 代码
try:
    # 获取苹果公司的 ticker 对象
    aapl = yf.Ticker('AAPL')
    # 获取多年的财务报表和股价数据
    financials = aapl.financials.T
    balance_sheet = aapl.balance_sheet.T
    info = aapl.info

    # 确保索引是 DatetimeIndex 以便合并
    financials.index = pd.to_datetime(financials.index)
    balance_sheet.index = pd.to_datetime(balance_sheet.index)

    # 提取所需数据
    df = pd.DataFrame()
    df['eps_basic'] = financials['Basic EPS']
    df['total_assets'] = balance_sheet['Total Assets']
    df['net_income'] = financials['Net Income']
    df['book_value'] = balance_sheet['Total Stockholder Equity']
    df['roa'] = df['net_income'] / df['total_assets']
    hist = aapl.history(period='max')['Close']
    df['pps'] = hist.resample('Y').last()
    df['market_cap'] = df['pps'] * info.get('sharesOutstanding', 1) # 添加备用值
    df['bm'] = df['book_value'] / df['market_cap']
    df.dropna(inplace=True)
    df = df.reset_index(drop=True)
except Exception as e:
    print(f"yfinance 数据获取失败，使用模拟数据代替。错误: {e}")
    # 创建一个结构相似的模拟 DataFrame
    data = {
        'eps_basic': [1.29, 1.91, 2.30, 2.98, 2.97, 3.28, 5.61, 6.15, 5.58, 5.95],
        'pps': [26.47, 39.04, 55.49, 41.31, 71.00, 131.14, 175.28, 155.74, 128.5, 190.1],
        'bm': [0.22, 0.16, 0.14, 0.20, 0.12, 0.05, 0.04, 0.04, 0.05, 0.03],
        'roa': [0.15, 0.16, 0.16, 0.19, 0.17, 0.21, 0.28, 0.29, 0.27, 0.28]
    }
    df = pd.DataFrame(data)

# 定义特征和目标
X = df[['pps', 'bm', 'roa']]
y = df['eps_basic']

print('数据集预览 (X):')
print(X.head())

X_train, X_test, y_train, y_test = train_test_split(
    X, y, 
    test_size=0.2, 
    random_state=42
)

print(f'训练集大小: {X_train.shape[0]} 条样本')
print(f'测试集大小: {X_test.shape[0]} 条样本')

# 1. 创建模型实例
rf_model = RandomForestRegressor(
    n_estimators=100, 
    random_state=42, 
    max_features='sqrt',
    oob_score=True
)

# 2. 在训练数据上拟合模型
rf_model.fit(X_train, y_train)

print(f'随机森林模型的 OOB 得分 (R²): {rf_model.oob_score_:.4f}')

# 3. 对测试集进行预测
y_pred_rf = rf_model.predict(X_test)

# 4. 计算均方误差
mse_rf = mean_squared_error(y_test, y_pred_rf)

print(f'随机森林模型的均方误差 (MSE): {mse_rf:.4f}')

# 1. 创建模型实例
xgb_model = XGBRegressor(
    n_estimators=100, 
    learning_rate=0.1, 
    max_depth=3, 
    random_state=42,
    objective='reg:squarederror' # 指定目标函数
)

# 2. 在训练数据上拟合模型
xgb_model.fit(X_train, y_train)

# 3. 对测试集进行预测
y_pred_xgb = xgb_model.predict(X_test)

# 4. 计算均方误差
mse_xgb = mean_squared_error(y_test, y_pred_xgb)

print(f'XGBoost 模型的均方误差 (MSE): {mse_xgb:.4f}')

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 5), sharey=True)

# 随机森林
ax1.scatter(y_test, y_pred_rf, alpha=0.7, edgecolors='k', c='#3498db')
ax1.plot([y_test.min(), y_test.max()], [y_test.min(), y_test.max()], 'r--', lw=2)
ax1.set_xlabel('真实 EPS')
ax1.set_ylabel('预测 EPS')
ax1.set_title(f'随机森林 (MSE: {mse_rf:.4f})')

# XGBoost
ax2.scatter(y_test, y_pred_xgb, alpha=0.7, edgecolors='k', c='#2ecc71')
ax2.plot([y_test.min(), y_test.max()], [y_test.min(), y_test.max()], 'r--', lw=2)
ax2.set_xlabel('真实 EPS')
ax2.set_title(f'XGBoost (MSE: {mse_xgb:.4f})')

plt.suptitle('模型预测性能可视化', fontsize=16)
plt.tight_layout(rect=[0, 0.03, 1, 0.95])
plt.show()

# 获取特征重要性
importance_rf = rf_model.feature_importances_
importance_xgb = xgb_model.feature_importances_
features = X.columns

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 5))

# 随机森林
sns.barplot(x=importance_rf, y=features, ax=ax1, palette='Blues_r')
ax1.set_title('随机森林 特征重要性')
ax1.set_xlabel('重要性得分')

# XGBoost
sns.barplot(x=importance_xgb, y=features, ax=ax2, palette='Greens_r')
ax2.set_title('XGBoost 特征重要性')
ax2.set_xlabel('重要性得分')

plt.tight_layout()
plt.show()

# 1. 创建原始 DataFrame
data = {
    'x1': [1.76, 0.40, 0.98, 2.24],
    'x2': [1.87, -0.98, 0.95, -0.15],
    'y': [-0.10, 0.41, 0.14, 1.45]
}
original_df = pd.DataFrame(data)

# 2. 生成三个自助样本
print('--- 原始数据 ---')
print(original_df)

for i in range(3):
    # 使用 random_state 确保每次运行结果一致
    bootstrap_sample = original_df.sample(n=len(original_df), replace=True, random_state=i)
    print(f'\n--- 自助样本 {i+1} (random_state={i}) ---')
    print(bootstrap_sample)

# 使用上一题的数据
X_ex = original_df[['x1', 'x2']]
y_ex = original_df['y']

# 调整 n_estimators
n_estimators_list = [2, 10, 50]
print('--- 调整 n_estimators (max_features="sqrt") ---')
for n in n_estimators_list:
    model = RandomForestRegressor(n_estimators=n, random_state=42, max_features='sqrt')
    model.fit(X_ex, y_ex)
    preds = model.predict(X_ex)
    mse = mean_squared_error(y_ex, preds)
    print(f'树的数量 = {n:2d}, 训练集 MSE = {mse:.4f}')

# 调整 max_features
print('\n--- 调整 max_features (n_estimators=50) ---')
# 'sqrt' 对 2 个特征来说是 1，2 是所有特征
max_features_list = ['sqrt', 2] 
for mf in max_features_list:
    model = RandomForestRegressor(n_estimators=50, random_state=42, max_features=mf)
    model.fit(X_ex, y_ex)
    preds = model.predict(X_ex)
    mse = mean_squared_error(y_ex, preds)
    print(f'特征数量 = {mf}, 训练集 MSE = {mse:.4f}')

# 调整 learning_rate
learning_rates = [0.01, 0.1, 0.5]
print('--- 调整 learning_rate (max_depth=3) ---')
for lr in learning_rates:
    model = XGBRegressor(n_estimators=100, learning_rate=lr, max_depth=3, random_state=42)
    model.fit(X_ex, y_ex)
    preds = model.predict(X_ex)
    mse = mean_squared_error(y_ex, preds)
    print(f'学习率 = {lr:.2f}, 训练集 MSE = {mse:.4f}')

# 调整 max_depth
max_depths = [1, 3, 5]
print('\n--- 调整 max_depth (learning_rate=0.1) ---')
for md in max_depths:
    model = XGBRegressor(n_estimators=100, learning_rate=0.1, max_depth=md, random_state=42)
    model.fit(X_ex, y_ex)
    preds = model.predict(X_ex)
    mse = mean_squared_error(y_ex, preds)
    print(f'最大深度 = {md}, 训练集 MSE = {mse:.4f}')