Module preprocess 

Chronix Preprocessing Pipeline

Gap detection, interpolation, smoothing, resampling, and clock drift correction for time-series data — all operating on Arrow-compatible &[f64] / &[i64] arrays with zero-copy where possible.

Pipeline Order

Raw data → Clock drift correction → Gap detection → Interpolation → Smoothing → Resampling

Interpolation Methods

• Linear — linearly interpolated values between known points
• Nearest — nearest neighbor fill
• CatmullRom — Catmull-Rom spline (C1 local interpolation, requires ≥ 4 surrounding points)
• Zero — fill with 0.0
• Forward — last known value carried forward
• Nocb — next observation carried backward

Smoothing Filters

• ExponentialSmoother — single-pass EMA with configurable alpha
• MovingAverageSmoother — sliding window mean
• WeightedMovingAverage — custom weight vector

Re-exports

pub use arrow_adapters::arrow_interpolate;

pub use arrow_adapters::arrow_preprocess;

pub use arrow_adapters::arrow_smooth;

pub use arrow_adapters::ArrowPreprocessResult;

pub use auto_features::AutoFeatureConfig;

pub use auto_features::FeatureMatrix;

pub use decomposition::detect_period;

pub use decomposition::detect_period_with_threshold;

pub use decomposition::stl_decompose;

pub use decomposition::Decomposition;

pub use decomposition::DecompositionError;

pub use decomposition::SeasonalDecomposer;

pub use decomposition::StlConfig;

pub use decomposition::StlDecomposer;

pub use features::diff;

pub use features::ewm;

pub use features::lag;

pub use features::pct_change;

pub use features::rolling_corr;

pub use features::rolling_mean;

pub use features::rolling_std;

pub use features::zscore;

Modules

arrow_adapters

Arrow-native wrappers for preprocessing operations.

auto_features

Automatic feature generation for time-series data.

decomposition

Seasonal-Trend decomposition using Loess (STL).

features

Temporal feature extraction functions for time-series data.

Structs

ClockDriftDetector

Detects and corrects clock drift in timestamp arrays.

DriftReport

Report of clock drift detection and correction.

ExponentialSmoother

Exponential smoothing with configurable alpha (0..1).

GapDetector

Detects gaps in a sorted timestamp column where the interval exceeds expected_interval × tolerance.

MovingAverageSmoother

Simple moving average with configurable window size.

PreprocessConfig

Configuration for the preprocessing pipeline.

PreprocessPipeline

Preprocessing pipeline.

PreprocessResult

Result of a preprocessing pipeline run.

ResampleConfig

Configuration for resampling.

Resampler

Resampler converts time-series data from one interval to another.

WeightedMovingAverage

Weighted moving average with custom weight vector.

Enums

AggregationFn

Aggregation function for downsampling.

ClockDriftStrategy

Strategy for clock drift correction.

InterpolationError

Errors that can occur during interpolation.

Interpolator

Interpolation strategy for gap filling.

PreprocessError

Preprocessing errors.

Smoother

Smoothing method enum for configuration.