Welcome to spotforecast2-safe (Core)

Version EU AI Act Audit Python

spotforecast2-safe is a specialized, hardened Python package for time series forecasting in safety-critical production environments. It provides a minimal, auditable core for feature engineering and recursive forecasting.

Software Identification (CPE)

For vulnerability tracking, supply chain management, and SBOM (Software Bill of Materials) generation:

cpe:2.3:a:sequential_parameter_optimization:spotforecast2_safe:*:*:*:*:*:*:*:*

Current release identifier:

cpe:2.3:a:sequential_parameter_optimization:spotforecast2_safe:0.30.10:*:*:*:*:*:*:*

See Model/Method Card for additional compliance details.

Installation

git clone https://github.com/sequential-parameter-optimization/spotforecast2-safe.git
cd spotforecast2-safe
uv sync

Safety-Critical Features

  • Zero Dead Code β€” No GUI, plotting, or AutoML dependencies (No Plotly, No Optuna).
  • Deterministic Transformations β€” Mathematical logic that ensures bit-level reproducibility.
  • Fail-Safe Processing β€” Explicit failure on dirty or incomplete data (NaNs/Infs) instead of silent imputation.
  • Minimal Footprint β€” Reduced attack surface for high-security deployment targets.

Core Capabilities

  • Data Service β€” Robust fetching of time series, weather, and holiday data.
  • Preprocessing β€” Hardened tools for data curation, resampling, temporal splitting, and Data Imputation & Gap Weighting.
  • Forecasting Engine β€” Simplified recursive forecasting and seasonal baselines.
Disclaimer & Liability

IMPORTANT: This software is provided β€œas is” and any express or implied warranties are disclaimed. The use of this software in safety-critical systems is at the sole risk of the user. For full details, see the Disclaimer in the Model Card.

Attributions

Parts of the code are ported from skforecast to reduce external dependencies. Many thanks to the skforecast team for their great work!