📈 Release: tseda (Time Series Explorer)
tseda is a specialized tool for the automated exploration and decomposition of regularly sampled time series data (hourly frequency or greater). It leverages Singular Spectral Analysis (SSA) to separate signals from noise without the manual guesswork of traditional modeling.
✨ What’s New
- Three-Step Workflow: A guided interface for initial assessment (KDE, ACF/PACF), automated SSA decomposition, and observation logging.
- Smart Automation:
- Dataset Suitability: Automatic check to identify structured data vs. white noise before you begin.
- Heuristic Grouping: Automatic component grouping based on the Durbin-Watson statistic to ensure defensible results.
- Change Point Detection: Built-in PELT algorithm to automatically identify shifts in trend and seasonal amplitude.
- Notebook & UI Parity: Full API support to run analyses in Python notebooks or via the interactive Dash web interface.
- Validated Export: Download Trend, Seasonality, and Noise components directly to CSV once grouping criteria are met.
🛠 Installation & Usage
Ensure you are using Python 3.13+.
Install the package
pip install tseda
Launch the interactive explorer
tseda
📋 Requirements
- Python 3.13 or higher.
- Designed for regularly sampled data (e.g., hourly, daily, monthly).
Would you like to include a link to a demo video or a specific technical breakdown of the SSA implementation in these notes?