时序预测之算法介绍

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_absolute_error, mean_squared_error

# 1. 模拟时序数据
print("=== 1. 模拟时序数据 ===")
np.random.seed(42)

dates = pd.date_range('2024-01-01', periods=7*24, freq='h')
power_load = []

for date in dates:
    hour = date.hour
    dayofweek = date.dayofweek
    month = date.month
    
    # 基础负荷
    if 0 <= hour < 6:
        base_load = 500
    elif 6 <= hour < 9:
        base_load = 800
    elif 9 <= hour < 17:
        base_load = 1000
    elif 17 <= hour < 21:
        base_load = 1200
    else:
        base_load = 700
    
    # 周末调整
    if dayofweek >= 5:
        base_load *= 0.8
    
    # 季节调整
    if month in [12, 1, 2]:  # 冬季
        base_load *= 1.2
    elif month in [6, 7, 8]:  # 夏季
        base_load *= 1.1
    
    load = base_load + np.random.normal(0, 50)
    power_load.append(max(100, load))

data = pd.DataFrame({
    'time': dates,
    'power_load': power_load
})

print(f"数据形状: {data.shape}")

# 2. 特征工程
print("\n=== 2. 特征工程 ===")

def extract_features(data):
    """从时间提取特征"""
    data = data.copy()
    
    # 基础时间特征
    data['hour'] = data['time'].dt.hour
    data['day'] = data['time'].dt.day
    data['month'] = data['time'].dt.month
    data['dayofweek'] = data['time'].dt.dayofweek
    data['dayofyear'] = data['time'].dt.dayofyear
    
    # 衍生特征
    data['is_weekend'] = data['dayofweek'].isin([5, 6]).astype(int)
    
    # 时间段
    data['time_period'] = pd.cut(
        data['hour'],
        bins=[-1, 5, 11, 13, 17, 23],
        labels=['凌晨', '上午', '中午', '下午', '晚上']
    )
    
    # 白天/晚上
    data['is_daytime'] = ((data['hour'] >= 6) & (data['hour'] < 18)).astype(int)
    
    # 季度
    data['quarter'] = data['time'].dt.quarter
    
    # 季节
    data['season'] = data['month'].map({
        12: '冬季', 1: '冬季', 2: '冬季',
        3: '春季', 4: '春季', 5: '春季',
        6: '夏季', 7: '夏季', 8: '夏季',
        9: '秋季', 10: '秋季', 11: '秋季'
    })
    
    return data

data_with_features = extract_features(data)
print(f"特征数: {data_with_features.shape[1]}")
print(data_with_features.head())

# 3. 准备训练数据
print("\n=== 3. 准备训练数据 ===")

def prepare_data(data, target='power_load'):
    """准备训练数据"""
    # 选择特征
    features = ['hour', 'day', 'month', 'dayofweek', 'is_weekend', 'is_daytime', 'quarter']
    X = data[features].values
    y = data[target].values
    
    return X, y, features

X, y, features = prepare_data(data_with_features)
print(f"特征: {features}")
print(f"X形状: {X.shape}")
print(f"y形状: {y.shape}")

# 4. 划分训练测试集
print("\n=== 4. 划分训练测试集 ===")
split_idx = int(len(X) * 0.8)
X_train, X_test = X[:split_idx], X[split_idx:]
y_train, y_test = y[:split_idx], y[split_idx:]

print(f"训练集: {X_train.shape}")
print(f"测试集: {X_test.shape}")

# 5. 线性回归
print("\n=== 5. 线性回归 ===")

def train_linear_regression(X_train, y_train, X_test, y_test):
    """训练线性回归模型"""
    model = LinearRegression()
    model.fit(X_train, y_train)
    
    y_pred = model.predict(X_test)
    
    mae = mean_absolute_error(y_test, y_pred)
    mse = mean_squared_error(y_test, y_pred)
    rmse = np.sqrt(mse)
    
    print(f"MAE: {mae:.4f}")
    print(f"MSE: {mse:.4f}")
    print(f"RMSE: {rmse:.4f}")
    
    return model, y_pred

lr_model, lr_pred = train_linear_regression(X_train, y_train, X_test, y_test)

# 6. 随机森林
print("\n=== 6. 随机森林 ===")

def train_random_forest(X_train, y_train, X_test, y_test):
    """训练随机森林模型"""
    model = RandomForestRegressor(
        n_estimators=100,
        max_depth=10,
        random_state=42
    )
    model.fit(X_train, y_train)
    
    y_pred = model.predict(X_test)
    
    mae = mean_absolute_error(y_test, y_pred)
    mse = mean_squared_error(y_test, y_pred)
    rmse = np.sqrt(mse)
    
    print(f"MAE: {mae:.4f}")
    print(f"MSE: {mse:.4f}")
    print(f"RMSE: {rmse:.4f}")
    
    # 特征重要性
    importances = model.feature_importances_
    for feat, imp in zip(features, importances):
        print(f"  {feat}: {imp:.4f}")
    
    return model, y_pred

rf_model, rf_pred = train_random_forest(X_train, y_train, X_test, y_test)

# 7. XGBoost（模拟）
print("\n=== 7. XGBoost（模拟）===")

def train_xgboost_simulated(X_train, y_train, X_test, y_test):
    """训练XGBoost模型（用GBDT模拟）"""
    from sklearn.ensemble import GradientBoostingRegressor
    
    model = GradientBoostingRegressor(
        n_estimators=100,
        max_depth=6,
        learning_rate=0.1,
        random_state=42
    )
    model.fit(X_train, y_train)
    
    y_pred = model.predict(X_test)
    
    mae = mean_absolute_error(y_test, y_pred)
    mse = mean_squared_error(y_test, y_pred)
    rmse = np.sqrt(mse)
    
    print(f"MAE: {mae:.4f}")
    print(f"MSE: {mse:.4f}")
    print(f"RMSE: {rmse:.4f}")
    
    return model, y_pred

xgb_model, xgb_pred = train_xgboost_simulated(X_train, y_train, X_test, y_test)

# 8. 模型对比
print("\n=== 8. 模型对比 ===")

def compare_models(y_test, predictions, names):
    """对比模型"""
    print("\n模型对比:")
    for name, pred in zip(names, predictions):
        mae = mean_absolute_error(y_test, pred)
        rmse = np.sqrt(mean_squared_error(y_test, pred))
        print(f"  {name}: MAE={mae:.4f}, RMSE={rmse:.4f}")

compare_models(y_test, [lr_pred, rf_pred, xgb_pred], 
               ['线性回归', '随机森林', 'XGBoost'])

# 9. 可视化预测结果
print("\n=== 9. 可视化预测结果 ===")
plt.figure(figsize=(14, 6))
plt.plot(y_test, label='真实值', marker='o', alpha=0.7)
plt.plot(lr_pred, label='线性回归', alpha=0.7)
plt.plot(rf_pred, label='随机森林', alpha=0.7)
plt.plot(xgb_pred, label='XGBoost', alpha=0.7)
plt.xlabel('时间步')
plt.ylabel('电力负荷')
plt.title('模型预测对比')
plt.legend()
plt.grid(True, alpha=0.3)
plt.savefig('model_comparison.png', dpi=100)
print("模型对比图已保存")

# 10. 算法选择说明
print("\n=== 10. 算法选择说明 ===")
print("""
时序预测算法选择:

1. 传统统计学算法
   - 简单平均
   - 指数平滑
   - AR、MA
   特点: 简单，但效果一般

2. 深度学习算法
   - RNN类算法
   - Transformer
   特点: 适合时序，但复杂

3. 传统机器学习算法
   - 线性回归
   - 决策树
   - 随机森林
   - SVM
   - LightGBM
   - XGBoost（本课使用）
   特点: 平衡效果和复杂度

本课选择: XGBoost
- 机器学习王牌
- 效率高
- Boosting思想
- 打分函数决定分支
""")

# 11. 特征工程说明
print("\n=== 11. 特征工程说明 ===")
print("""
特征工程（从时间提取特征）:

1. 基础时间特征
   - hour: 小时
   - day: 日
   - month: 月
   - dayofweek: 星期几

2. 衍生特征
   - is_weekend: 是否周末
   - time_period: 时间段（凌晨/上午/中午/下午/晚上）
   - is_daytime: 是否白天
   - quarter: 季度
   - season: 季节

作用: 一个时间转成多个特征变量
      变成多变量单步预测
""")

# 12. 完整流程
def time_series_algorithm_pipeline():
    """时序预测算法完整流程"""
    print("=" * 50)
    print("时序预测算法 - 完整流程")
    print("=" * 50)
    
    print("\n1. 算法选择")
    print("   传统统计学、深度学习、传统机器学习")
    print("   本课选择: XGBoost")
    
    print("\n2. 特征工程")
    print("   从时间提取多个特征变量")
    print("   转成二维数据结构（DataFrame）")
    
    print("\n3. 模型训练")
    print("   划分训练测试集")
    print("   训练XGBoost模型")
    
    print("\n4. 模型评估")
    print("   MAE、MSE、RMSE")
    
    print("\n" + "=" * 50)
    print("算法介绍完成!")
    print("=" * 50)

time_series_algorithm_pipeline()