Key Concepts
Understanding the foundations of time series forecasting with TimeGPT
These key concepts cover the foundations of time series data, how forecasts are generated, and the role of TimeGPT in predicting future values and detecting anomalies.
Use these concepts as a reference to better understand how TimeGPT simplifies tasks such as demand forecasting, anomaly detection, and multi-series forecasting.
Time Series
A sequence of numerical data points arranged in chronological order.
Forecasting
Predicting future values by analyzing historical data and patterns.
Anomaly Detection
Identifying unusual or unexpected events that deviate from typical behavior.
Multiple Series
Managing and forecasting multiple time series data at once.
TimeGPT
Nixtla’s generative pre-trained model for time series forecasting.
Inputs (Tokens)
Segments of historical data that inform TimeGPT’s forecasting process.
Using TimeGPT
Below is a simple, high-level workflow to help you get started with TimeGPT for your forecasting tasks.
Prepare Your Dataset
Ensure your time series data is clean and in a consistent format. Identify any exogenous variables (like holidays, promotions, or economic indicators) that might affect the forecast.
Configure TimeGPT Parameters
Determine the forecast horizon (how far into the future you want to predict) and other model settings or hyperparameters if applicable.
Request a Forecast
In practice, you may use Python, R, or another language to call the TimeGPT API. Provide your historical data and specify any additional parameters required by the endpoint.
Evaluate the Results
Review the returned forecast values. Compare these predictions with actual outcomes over time to assess accuracy, and fine-tune your approach or data quality as needed.
Congratulations! You’re now ready to use TimeGPT for more accurate forecasting, streamlined anomaly detection, and multi-series analysis.