时序数据分类介绍

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

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

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

for date in dates:
    hour = date.hour
    dayofweek = date.dayofweek
    
    # 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
    
    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}")
print(data.head())

# 2. 单变量单步预测
print("\n=== 2. 单变量单步预测 ===")

def single_variable_single_step(data, train_size=0.8):
    """单变量单步预测"""
    print("\n--- 单变量单步 ---")
    
    # 准备数据
    X = data[['power_load']].values[:-1]  # 特征（历史）
    y = data['power_load'].values[1:]    # 标签（未来一步）
    
    # 划分训练测试集
    split_idx = int(len(X) * train_size)
    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}")
    
    # 使用线性回归模拟（实际用XGBoost）
    from sklearn.linear_model import LinearRegression
    model = LinearRegression()
    model.fit(X_train, y_train)
    
    # 预测
    y_pred = model.predict(X_test)
    
    # 评估
    from sklearn.metrics import mean_absolute_error, mean_squared_error
    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}")
    
    # 可视化
    plt.figure(figsize=(12, 6))
    plt.plot(y_test, label='真实值', marker='o')
    plt.plot(y_pred, label='预测值', marker='x')
    plt.xlabel('时间步')
    plt.ylabel('电力负荷')
    plt.title('单变量单步预测')
    plt.legend()
    plt.grid(True)
    plt.savefig('single_single_step.png', dpi=100)
    
    return model

model_single = single_variable_single_step(data)

# 3. 单变量多步预测（滚动预测）
print("\n=== 3. 单变量多步预测（滚动预测）===")

def single_variable_multi_step(data, n_steps=24):
    """单变量多步预测（滚动预测）"""
    print("\n--- 单变量多步（滚动预测）---")
    
    # 准备数据
    X = data[['power_load']].values[:-1]
    y = data['power_load'].values[1:]
    
    # 训练模型
    from sklearn.linear_model import LinearRegression
    model = LinearRegression()
    model.fit(X, y)
    
    # 滚动预测
    print(f"预测未来{n_steps}个时间步")
    
    # 最后一个已知值
    last_value = data['power_load'].values[-1]
    predictions = []
    current_value = last_value
    
    for step in range(n_steps):
        # 预测下一个时间步
        next_pred = model.predict([[current_value]])[0]
        predictions.append(next_pred)
        # 更新当前值（滚动）
        current_value = next_pred
        print(f"  时间步{step+1}: 预测值={next_pred:.2f}")
    
    # 可视化
    plt.figure(figsize=(12, 6))
    
    # 历史数据
    history = data['power_load'].values[-48:]  # 最后48个点
    plt.plot(range(len(history)), history, 'b-', label='历史数据')
    
    # 预测数据
    future_x = range(len(history)-1, len(history)-1+n_steps)
    plt.plot(future_x, predictions, 'r--', label='预测数据', marker='o')
    
    plt.xlabel('时间步')
    plt.ylabel('电力负荷')
    plt.title('单变量多步预测（滚动预测）')
    plt.legend()
    plt.grid(True)
    plt.savefig('single_multi_step.png', dpi=100)
    
    return predictions

predictions = single_variable_multi_step(data, n_steps=24)

# 4. 多变量单步预测
print("\n=== 4. 多变量单步预测 ===")

def multi_variable_single_step(data, train_size=0.8):
    """多变量单步预测"""
    print("\n--- 多变量单步 ---")
    
    # 提取时间特征
    data = data.copy()
    data['hour'] = data['time'].dt.hour
    data['dayofweek'] = data['time'].dt.dayofweek
    data['is_weekend'] = data['dayofweek'].isin([5, 6]).astype(int)
    
    # 准备数据
    features = ['power_load', 'hour', 'dayofweek', 'is_weekend']
    X = data[features].values[:-1]
    y = data['power_load'].values[1:]
    
    # 划分训练测试集
    split_idx = int(len(X) * train_size)
    X_train, X_test = X[:split_idx], X[split_idx:]
    y_train, y_test = y[:split_idx], y[split_idx:]
    
    print(f"特征数: {len(features)}")
    print(f"训练集: {X_train.shape}")
    print(f"测试集: {X_test.shape}")
    
    # 训练
    from sklearn.linear_model import LinearRegression
    model = LinearRegression()
    model.fit(X_train, y_train)
    
    # 预测
    y_pred = model.predict(X_test)
    
    # 评估
    from sklearn.metrics import mean_absolute_error
    mae = mean_absolute_error(y_test, y_pred)
    print(f"MAE: {mae:.4f}")
    
    return model

model_multi = multi_variable_single_step(data)

# 5. 四大分类对比
print("\n=== 5. 四大分类对比 ===")

def compare_categories():
    """四大分类对比"""
    print("""
时序预测四大分类:

1. 单变量单步
   - 特征: 1列
   - 预测: 1个时间步
   - 方法: 直接建模预测
   - 难度: 简单

2. 单变量多步
   - 特征: 1列
   - 预测: 多个时间步
   - 方法: 建模 + 滚动预测
   - 难度: 中等

3. 多变量单步
   - 特征: 多列
   - 预测: 1个时间步
   - 方法: 直接建模预测
   - 难度: 中等

4. 多变量多步
   - 特征: 多列
   - 预测: 多个时间步
   - 方法: 建模 + 滚动预测
   - 难度: 复杂
""")

compare_categories()

# 6. 滚动预测演示
print("\n=== 6. 滚动预测演示 ===")

def rolling_prediction_demo(data, n_days=3):
    """滚动预测演示"""
    print("\n--- 滚动预测演示 ---")
    
    # 训练模型
    X = data[['power_load']].values[:-1]
    y = data['power_load'].values[1:]
    
    from sklearn.linear_model import LinearRegression
    model = LinearRegression()
    model.fit(X, y)
    
    # 模拟滚动预测
    print("模拟3天滚动预测:")
    
    current_value = data['power_load'].values[-1]
    
    for day in range(n_days):
        print(f"\n第{day+1}天:")
        daily_predictions = []
        
        for hour in range(24):
            next_pred = model.predict([[current_value]])[0]
            daily_predictions.append(next_pred)
            current_value = next_pred
        
        print(f"  预测24小时")
        print(f"  最高负荷: {max(daily_predictions):.2f}")
        print(f"  最低负荷: {min(daily_predictions):.2f}")
        print(f"  平均负荷: {np.mean(daily_predictions):.2f}")

rolling_prediction_demo(data, n_days=3)

# 7. 完整流程
def time_series_classification():
    """时序数据分类介绍"""
    print("=" * 50)
    print("时序数据分类介绍")
    print("=" * 50)
    
    print("\n1. 四大分类")
    print("   单变量单步、单变量多步")
    print("   多变量单步、多变量多步")
    
    print("\n2. 单变量单步")
    print("   一列特征，预测一个时间步")
    print("   直接建模预测")
    
    print("\n3. 单变量多步")
    print("   一列特征，预测多个时间步")
    print("   建模 + 滚动预测")
    
    print("\n4. 滚动预测")
    print("   时间不断增长")
    print("   预测值变成新的历史数据")
    
    print("\n5. 建模算法")
    print("   XGBoost（极限梯度提升树）")
    
    print("\n" + "=" * 50)
    print("分类介绍完成!")
    print("=" * 50)

time_series_classification()