sdes

Input Stack: beta: Double alpha: Double training: Int expr: TimeSeriesExpr

⇨

Output Stack: TimeSeriesExpr

Sliding double exponential smoothing - a deterministic variant of :des that provides consistent predictions regardless of when data feeding starts. This operator is recommended over standard DES for alerting and analysis where reproducible results are essential.

Parameters

• expr: The time series expression to apply smoothing to
• training: Number of data points to use for initial training before generating predictions
• alpha: Smoothing factor for the level component (0.0 to 1.0, higher = more responsive to recent changes)
• beta: Smoothing factor for the trend component (0.0 to 1.0, higher = more responsive to recent trend changes)

How It Works

SDES solves the determinism problem of standard DES by using two alternating DES functions:

1. While one function is training on new data, the other provides predictions
2. After the training window, roles are swapped
3. This ensures predictions for any given time are always the same

Execution Timeline

 F1 | A |-- T1 --|-- P1 --|-- T1 --|-- P1 --|-- T1 --|
 F2 | A |        |-- T2 --|-- P2 --|-- T2 --|-- P2 --|

Result:
 R  |-- NaN -----|-- P1 --|-- P2 --|-- P1 --|-- P2 --|

Where:

• A: Both functions wait for training window boundary
• T1/T2: Training phases for each function
• P1/P2: Prediction phases alternating between functions
• NaN: No predictions until first training completes

Advantages Over Standard DES

1. Deterministic: Same input always produces same output
2. Alerting-friendly: Consistent thresholds for alarm tuning
3. Faster adaptation: Responds more quickly to sharp changes

Visual Characteristics

The alternation between functions can create visual differences compared to standard DES:

Gradual changes may appear choppier:

Sharp changes are detected faster:

Examples

Applying SDES with 5-point training, alpha=0.1, beta=0.5:

Before	After
name,requestsPerSecond,:eq, :sum, :per-step	name,requestsPerSecond,:eq, :sum, 5,0.1,0.5,:sdes

Related Operations

• :des - Standard double exponential smoothing (non-deterministic)
• :rolling-mean - Simple moving average (simpler but less sophisticated)
• :trend - Deprecated moving average operator

Since: 1.5.0