​
Multiple Series Forecasting with TimeGPT

TimeGPT can concurrently forecast multiple series at once. To do this, you must provide a DataFrame with multiple unique values defined under the unique_id column.

TimeGPT is a powerful forecasting solution that supports simultaneous predictions for multiple time series. This guide will walk you through setting up your Nixtla Client, loading data, and generating forecasts.

​
How It Works

Key Concept

Forecasting multiple series requires each observation to have a unique identifier under the unique_id column. TimeGPT automatically handles each series individually and returns forecasts for every unique series in your dataset.

1

1. Import Required Libraries

import pandas as pd
from nixtla import NixtlaClient
2

2. Initialize the Nixtla Client

Choose between the default Nixtla endpoint or an Azure AI endpoint.

nixtla_client = NixtlaClient(
    # defaults to os.environ.get("NIXTLA_API_KEY")
    api_key='my_api_key_provided_by_nixtla'
)
3

3. Prepare Your Data

# Load data
df = pd.read_csv(
    'https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/electricity-short.csv'
)
4

4. Forecast Multiple Series

# Forecast multiple series
forecast_df = nixtla_client.forecast(
    df=df,
    h=24
)

Available models in Azure AI
If using an Azure AI endpoint, set the model parameter explicitly to "azureai":

nixtla_client.forecast(
    ...,
    model="azureai"
)

Choosing the Right Model

If you’re using the public API, two models are supported: timegpt-1 and timegpt-1-long-horizon.
The default is timegpt-1. Check out the long horizon tutorial to learn when and how to apply timegpt-1-long-horizon.

For further details, visit the detailed tutorial:
Multiple series forecasting.