RollingMean / Computation Layer

Calculate the rolling (moving) arithmetic mean over a sliding window. Similar to pandas.rolling().mean() or R's rollmean().

Mathematical form: $y_{i} = \frac{\sum _{j = i - w + 1}^{i} ( x _{j} \cdot w _{j} )}{\sum _{j = i - w + 1}^{i} w _{j}}$ where $w$ is the window size and $w_{j}$ are optional weights.

Common applications:

Technical analysis (moving averages)
Trend analysis (smoothing data)
Signal processing (noise reduction)
Time series analysis (seasonality removal)
Sales forecasting (trend identification)
Weather data analysis (temperature trends)
Performance monitoring (running averages)

Note: Float32/Float64 inputs produce corresponding precision outputs.

f32, f64

Table

f32, f64

Select

column

The numeric column to compute rolling means for. Typical inputs:

Price series (stock prices, commodity prices)
Measurement data (temperatures, pressures)
Time series metrics (daily sales, hourly traffic)
Sensor readings (continuous monitoring data)

WindowSize

u32

Number of observations in each sliding window. Common values:

5: Weekly average (business days)
20: Monthly average (trading days)
30: Monthly average (calendar days)
252: Annual average (trading days)

Larger windows provide more smoothing but lag trends more.

Weights

[f64, ...]

Optional weights for each window position. Must match window size if provided. Uses:

Exponential weighting (recent data more important)
Linear decay (gradual importance reduction)
Custom patterns (domain-specific weighting)

When omitted, equal weights are used (simple moving average).

Require complete windows for calculation. Ensures statistical reliability by only computing means when all values in window are available. Used when data completeness is critical for analysis integrity.

Period

Value

u32

Minimum number of valid observations (Value) required in window for calculation. Examples:

1: Calculate with any valid data
window_size/2: Require 50% data coverage
window_size: Require complete window

Balances data availability with statistical reliability.

Center

bool

false

Position of the window label. When true, label is at window center; when false, at window end. Effects:

true: Better alignment with trends, lag split between past and future
false: Only uses past data, better for real-time processing

AsColumn

name

Name for the new column. If not provided, the system generates a unique name. If AsColumn matches an existing column, the existing column is replaced. The name should follow valid column naming conventions.