What evaluation metrics are included in the results?

The skill automatically calculates SMAPE, MASE, MAE, and RMSE to help you compare model performance objectively.

What forecasting models does this skill support?

It supports a wide range of models including StatsForecast (AutoARIMA, AutoETS), MLForecast (XGBoost, LightGBM), and the foundation model TimeGPT.

Can I use this for real-time forecasting?

Yes, it scaffolds the experiment harness which can be adapted into a production service or used for recurring batch forecasts.

Do I need a Nixtla API key to use this skill?

An API key is only required if you intend to use TimeGPT; local models like StatsForecast and MLForecast work without one.

Nixtla Experiment Architect

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

byintent-solutions-io

Data Science & ML

Orchestrates production-ready forecasting experiments and model benchmarking using the Nixtla ecosystem.

About

The Nixtla Experiment Architect skill streamlines the creation of robust time-series forecasting pipelines by scaffolding complete experiment harnesses. It automates configuration management, enables multi-model comparisons between StatsForecast, MLForecast, and TimeGPT, and implements advanced cross-validation techniques like rolling-origin windows. This skill is essential for data scientists and developers looking to move from raw data to validated forecasting models with standardized metrics like SMAPE and MASE, ensuring production-grade reliability and performance.

Key Features

Comprehensive model benchmarking with built-in cross-validation
Seamless integration with Nixtla's cloud-based TimeGPT API
Automated experiment scaffolding for StatsForecast, MLForecast, and TimeGPT
YAML-based configuration management for reproducible forecasting runs
Standardized metrics evaluation including SMAPE, MASE, MAE, and RMSE
0 GitHub stars

Use Cases

Setting up automated backtesting and cross-validation pipelines for financial forecasting
Prototyping production-ready forecasting services using StatsForecast and TimeGPT
Benchmarking multiple time-series models to identify the most accurate predictor for sales or demand data

About

Key Features

Comprehensive model benchmarking with built-in cross-validation
Seamless integration with Nixtla's cloud-based TimeGPT API
Automated experiment scaffolding for StatsForecast, MLForecast, and TimeGPT
YAML-based configuration management for reproducible forecasting runs
Standardized metrics evaluation including SMAPE, MASE, MAE, and RMSE
0 GitHub stars

Use Cases

Setting up automated backtesting and cross-validation pipelines for financial forecasting
Prototyping production-ready forecasting services using StatsForecast and TimeGPT
Benchmarking multiple time-series models to identify the most accurate predictor for sales or demand data