What models are supported for cross-validation?

The skill supports Nixtla's TimeGPT, as well as statistical models from StatsForecast including AutoARIMA, AutoETS, AutoTheta, and SeasonalNaive.

How does this skill handle different data frequencies?

You can specify the frequency of your time series data using standard pandas frequency aliases (e.g., 'D' for daily, 'M' for monthly) during the execution step.

What metrics does the skill provide?

It generates a metrics.json file containing Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) across all cross-validation folds.

Do I need an API key to use this skill?

A NIXTLA_TIMEGPT_API_KEY is only required if you are using the TimeGPT model. Local statistical models like ARIMA or ETS do not require an API key.

What happens if my dataset is too small?

The skill includes built-in error handling that validates if your time series length is sufficient for the requested number of folds and window size.

Nixtla Cross-Validator

Name: Nixtla Cross-Validator
Author: intent-solutions-io

byintent-solutions-io

Ciencia de Datos y ML

Evaluates time series model performance using rigorous cross-validation techniques like expanding and sliding windows.

Acerca de

The Nixtla Cross-Validator skill enables Claude to perform professional-grade time series backtesting and model evaluation. By automating the split of historical data into training and validation sets, it allows users to assess how forecasting models like TimeGPT, ARIMA, and ETS generalize to unseen data. It calculates essential accuracy metrics such as MAE and RMSE and generates visual plots, ensuring data scientists and developers can confidently select the most robust models for their forecasting pipelines.

Características Principales

Integration with TimeGPT and StatsForecast (ARIMA, ETS, Theta)
Visual plotting of predicted vs. actual values for each fold
0 GitHub stars
Data sufficiency validation and automated window configuration
Automated expanding and sliding window cross-validation
Automatic calculation of error metrics including MAE and RMSE

Casos de Uso

Comparing TimeGPT performance against traditional statistical models
Backtesting financial or demand forecasting models on historical data
Automating model selection based on standardized performance metrics

Acerca de

Características Principales

Integration with TimeGPT and StatsForecast (ARIMA, ETS, Theta)
Visual plotting of predicted vs. actual values for each fold
0 GitHub stars
Data sufficiency validation and automated window configuration
Automated expanding and sliding window cross-validation
Automatic calculation of error metrics including MAE and RMSE

Casos de Uso

Comparing TimeGPT performance against traditional statistical models
Backtesting financial or demand forecasting models on historical data
Automating model selection based on standardized performance metrics