algo-forecast-exponential

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Exponential Smoothing

指数平滑法

Overview

概述

Exponential smoothing assigns exponentially decreasing weights to past observations. Three variants: Simple (SES, level only), Holt (level + trend), Holt-Winters (level + trend + seasonality). ETS framework (Error-Trend-Seasonality) provides a unified statistical model. Fast, interpretable, and competitive with complex models for short horizons.

指数平滑法为历史观测值分配呈指数递减的权重。包含三种变体：简单指数平滑（SES，仅考虑水平项）、Holt双参数指数平滑（水平项+趋势项）、Holt-Winters三参数指数平滑（水平项+趋势项+季节性项）。ETS框架（误差-趋势-季节性）提供了统一的统计模型。该方法运算快速、可解释性强，在短期预测场景下可与复杂模型相媲美。

When to Use

适用场景

Trigger conditions:

Quick forecasting with minimal configuration
Short-horizon forecasts (1-2 seasonal cycles ahead)
Data with clear level, trend, and/or seasonal components

When NOT to use:

For long-range forecasts (uncertainty accumulates too fast)
When external regressors are important (use regression or ML models)

触发条件：

只需极少配置的快速预测
短期预测（未来1-2个周期的季节性数据）
具有明显水平项、趋势项和/或季节性项的数据

不适用场景：

长期预测（不确定性累积过快）
外部回归因子很重要的场景（使用回归模型或机器学习模型）

Algorithm

算法

IRON LAW: Smoothing Parameters Control the Bias-Variance Trade-Off
α (level), β (trend), γ (seasonality) range [0,1].
- α near 1: react quickly to changes, noisy forecasts (high variance)
- α near 0: smooth forecasts, slow to adapt (high bias)
Optimize via minimizing MSE on training data (or use information criteria).
Never hand-pick smoothing parameters without validation.

IRON LAW: Smoothing Parameters Control the Bias-Variance Trade-Off
α (level), β (trend), γ (seasonality) range [0,1].
- α near 1: react quickly to changes, noisy forecasts (high variance)
- α near 0: smooth forecasts, slow to adapt (high bias)
Optimize via minimizing MSE on training data (or use information criteria).
Never hand-pick smoothing parameters without validation.

Phase 1: Input Validation

阶段1：输入验证

Identify components: level only (SES), level+trend (Holt), level+trend+seasonality (Holt-Winters). Determine: additive vs multiplicative trend/seasonality. Gate: Component structure identified, seasonal period known.

识别数据成分：仅水平项（SES）、水平项+趋势项（Holt）、水平项+趋势项+季节性项（Holt-Winters）。确定：加法型 vs 乘法型趋势/季节性。 检查点： 已识别成分结构，已知季节性周期。

Phase 2: Core Algorithm

阶段2：核心算法

Holt-Winters (additive):

Initialize: level₀ = mean(first season), trend₀ = (mean(season 2) - mean(season 1))/s, seasonal₀ from first season deviations
Update equations at each t:
- Level: ℓₜ = α(yₜ - sₜ₋ₛ) + (1-α)(ℓₜ₋₁ + bₜ₋₁)
- Trend: bₜ = β(ℓₜ - ℓₜ₋₁) + (1-β)bₜ₋₁
- Seasonal: sₜ = γ(yₜ - ℓₜ) + (1-γ)sₜ₋ₛ
Forecast: ŷₜ₊ₕ = ℓₜ + h×bₜ + sₜ₊ₕ₋ₛ

Holt-Winters（加法型）：

初始化：level₀ = 第一个周期的均值，trend₀ = (第二个周期均值 - 第一个周期均值)/s，seasonal₀ 来自第一个周期的偏差值
每个时间步t的更新公式：
- 水平项：ℓₜ = α(yₜ - sₜ₋ₛ) + (1-α)(ℓₜ₋₁ + bₜ₋₁)
- 趋势项：bₜ = β(ℓₜ - ℓₜ₋₁) + (1-β)bₜ₋₁
- 季节性项：sₜ = γ(yₜ - ℓₜ) + (1-γ)sₜ₋ₛ
预测：ŷₜ₊ₕ = ℓₜ + h×bₜ + sₜ₊ₕ₋ₛ

Phase 3: Verification

阶段3：验证

Check: in-sample RMSE, residual patterns. Compare against naive baselines (last value, seasonal naive). Gate: Beats naive baseline, residuals show no systematic pattern.

检查：样本内RMSE、残差模式。与朴素基线（最后一个值、季节性朴素法）对比。 检查点： 性能优于朴素基线，残差无系统性模式。

Phase 4: Output

阶段4：输出

Return forecasts with smoothed components.

返回包含平滑成分的预测结果。

Output Format

输出格式

json

{
  "forecasts": [{"period": "2025-04", "forecast": 1150, "level": 1100, "trend": 20, "seasonal": 30}],
  "parameters": {"alpha": 0.3, "beta": 0.1, "gamma": 0.15},
  "metadata": {"method": "holt_winters_additive", "seasonal_period": 12, "rmse": 45}
}

json

{
  "forecasts": [{"period": "2025-04", "forecast": 1150, "level": 1100, "trend": 20, "seasonal": 30}],
  "parameters": {"alpha": 0.3, "beta": 0.1, "gamma": 0.15},
  "metadata": {"method": "holt_winters_additive", "seasonal_period": 12, "rmse": 45}
}

Examples

示例

Sample I/O

输入输出样例

Input: 36 months of monthly sales, clear upward trend, December spike Expected: Holt-Winters additive. Forecast continues trend with repeated December seasonality.

输入： 36个月的月度销售数据，有明显上升趋势，12月出现峰值 预期结果： 使用Holt-Winters加法型。预测结果延续趋势并重复12月的季节性特征。

Edge Cases

边缘情况

Input	Expected	Why
No trend, no seasonality	SES (α only)	Simplest variant suffices
Seasonal amplitude grows	Use multiplicative	Additive would underestimate peaks
Very short series (<2 seasons)	SES or Holt only	Can't estimate seasonality

输入	预期结果	原因
无趋势、无季节性	SES（仅α参数）	最简单的变体已足够
季节性幅度随水平增长	使用乘法型	加法型会低估峰值
极短序列（<2个周期）	SES或仅Holt	无法估计季节性

Gotchas

注意事项

Additive vs multiplicative: If seasonal swings grow proportionally with level, use multiplicative. Wrong choice produces poor forecasts, especially at extremes.
Initialization sensitivity: The first season's values set the baseline. Poor initialization from noisy early data propagates through the entire forecast.
Damped trend: For long horizons, linear trend extrapolation is unrealistic. Use damped trend (φ parameter) to flatten the trend over time.
Multiple seasonalities: Standard Holt-Winters handles one seasonal period. For daily data with weekly AND yearly patterns, use TBATS or STL+ETS.
Outlier sensitivity: A single outlier can shift the level estimate significantly (especially with high α). Pre-detect and handle outliers.

加法型vs乘法型：如果季节性波动随水平成比例增长，使用乘法型。选择错误会导致预测效果差，尤其是在极端值时。
初始化敏感性：第一个周期的值设定了基线。早期噪声数据导致的初始化不当会影响整个预测过程。
阻尼趋势：对于长期预测，线性趋势外推不现实。使用阻尼趋势（φ参数）使趋势随时间趋平。
多季节性：标准Holt-Winters仅处理一个季节性周期。对于同时具有周度和年度模式的日度数据，使用TBATS或STL+ETS。
异常值敏感性：单个异常值会显著改变水平估计（尤其是当α值较高时）。需预先检测并处理异常值。

References

参考资料

For ETS framework and model selection, see
```
references/ets-framework.md
```
For damped trend variants, see
```
references/damped-trend.md
```

关于ETS框架和模型选择，参见
```
references/ets-framework.md
```
关于阻尼趋势变体，参见
```
references/damped-trend.md
```