DeTime

DeTime is a time-series decomposition Python package and CLI workflow layer for trend, oscillation, residual, method-specific components, and metadata across univariate and aligned multichannel decomposition workflows.

Scientific toolkit / workflow system

Decompose time series without redesigning the workflow.

One interface for trend, oscillation, residual, and metadata.

Give DeTime one series or aligned multichannel data. It returns trend, seasonal or oscillatory structure, residuals, method-specific components, and metadata through the same Python and CLI interface.

Run First Example Install from GitHub

STL Decomposition workflow-oriented time-series decomposition

Observed

Trend

Seasonal

Residual

Decomposition complete

Observed

Trend

Residual

Python and CLI entrypointsStable commands for decomposition workflows.
Flagship method supportSSASTDSTDRMSSA
Published examplesReal stdout, plots, and saved artifacts.
Machine-facing schemasRecommendation and low-token result modes for automation.

Canonical import: detime Distribution: de-time Hugging Face mirror: Zipeng365/DeTime Flagship methods: SSA / STD / STDR / MSSA Machine-facing schemas and low-token result modes

Quant Trading Tutorials

New tutorial column

Turn market series into audited decomposition features.

The quant trading column keeps DeTime in its proper role: a decomposition layer that routes real market inputs into trend, cycle, residual, signal, and backtest research steps.

10 notebooks for data loading, timing, pairs, factor selection, rotation, adapters, and audit Real runtime market data policy for US, Korea, ETF, and crypto examples Walk-forward validation before signal evaluation

Open Quant Column Strategy Map

DeTime Research STL / Walk-forward audit ready

Universe Market inputs SPY 005930.KS BTC-USD real data only

Observed price

Trend

Cycle

Residual

Notebook path Features -> Signals -> Backtest 02 Timing vectorbt 04 Pairs residual 09 Validation audit

Why DeTime exists

Reason for the tool

A stable workflow layer for time-series decomposition.

DeTime exists because decomposition work often moves between notebooks, method-specific wrappers, CLI scripts, and machine-facing automation. The package keeps the method choice flexible while preserving one Python/CLI surface and one result contract.

Different decomposition methods expose different interfaces One decompose() entrypoint

Results are hard to compare One DecompResult for trend, season, residual, components, and meta

CLI and Python workflows often split One DecompositionConfig model across Python and CLI usage

Automation needs compact outputs Schemas, recommendations, and low-token result views

Data in, components out

Workflow

Data in, components out

DeTime keeps the user-facing contract stable while the method underneath can change. The same shape of result comes back whether you start with a single series or an aligned multichannel panel.

1D series or aligned 2D panel DecompositionConfig(method, params) decompose(...) or detime run trend, season, residual, components, metadata

Input 1D series or aligned 2D panel

Config DecompositionConfig(method, params)

Run decompose(...) or detime run

Output trend, season, residual, components, metadata