What types of cross-validation are supported?

The skill scaffolds both rolling-origin and expanding-window cross-validation workflows for robust model evaluation.

Can I customize the evaluation metrics?

Yes, the generated harness defaults to SMAPE, MASE, and RMSE, but the configuration can be extended to include other standard Nixtla metrics.

Do I need an API key to use this skill?

While statistical and ML models work locally, using the TimeGPT features requires a valid NIXTLA_API_KEY from Nixtla.

Which Nixtla libraries does this skill support?

It supports StatsForecast, MLForecast, and TimeGPT, providing a unified harness for statistical, machine learning, and foundation models.

Nixtla Experiment Architect

Name: Nixtla Experiment Architect
Author: intent-solutions-io

byintent-solutions-io

Ciencia de Datos y ML

Scaffolds production-ready time-series forecasting experiments using Nixtla's suite of machine learning and statistical libraries.

Acerca de

This skill automates the setup of comprehensive forecasting pipelines by generating configuration files, experiment harnesses, and evaluation workflows. It streamlines the comparison of diverse models—including StatsForecast, MLForecast, and TimeGPT—while implementing robust cross-validation techniques like rolling-origin or expanding-window. It is ideal for data scientists and developers who need to quickly move from raw time-series data to a structured, reproducible benchmarking environment with standardized performance metrics like SMAPE and MASE.

Características Principales

Comprehensive metrics evaluation including SMAPE, MASE, and RMSE
0 GitHub stars
Multi-model comparison across StatsForecast, MLForecast, and TimeGPT
Standardized cross-validation (rolling-origin and expanding-window)
Automated YAML-based experiment configuration generation
Scaffolding of runnable Python experiment harnesses

Casos de Uso

Setting up automated hourly energy consumption prediction pipelines
Evaluating TimeGPT foundation model performance against traditional statistical methods
Benchmarking multiple forecasting models against historical business sales data

Acerca de

Características Principales

Comprehensive metrics evaluation including SMAPE, MASE, and RMSE
0 GitHub stars
Multi-model comparison across StatsForecast, MLForecast, and TimeGPT
Standardized cross-validation (rolling-origin and expanding-window)
Automated YAML-based experiment configuration generation
Scaffolding of runnable Python experiment harnesses

Casos de Uso

Setting up automated hourly energy consumption prediction pipelines
Evaluating TimeGPT foundation model performance against traditional statistical methods
Benchmarking multiple forecasting models against historical business sales data