TimeGPT-1

The first foundation model for time series forecasting and anomaly detection

TimeGPT is a production-ready, generative pretrained transformer for time series forecasting, developed by Nixtla. It is capable of accurately predicting various domains such as retail, electricity, finance, and IoT, with just a few lines of code. Additionally, it can detect anomalies in time series data.

TimeGPT was initially developed in Python but is now available to R users through thenixtlar package.

Table of Contents

• Installation
• Forecast Using TimeGPT in 3 Easy Steps
• Anomaly Detection Using TimeGPT in 3 Easy Steps
• Features and Capabilities
• Documentation
• API Support
• How to Cite
• License
• Get in Touch

Installation

nixtlar is available on CRAN, so you can install the latest stable version usinginstall.packages.

# Install nixtlar from CRANinstall.packages("nixtlar")# Then load itlibrary(nixtlar)

Alternatively, you can install the development version ofnixtlar fromGitHub withdevtools::install_github.

# install.packages("devtools")devtools::install_github("Nixtla/nixtlar")

Forecast Using TimeGPT in 3 Easy Steps

library(nixtlar)

1. Set your API key. Get yours atdashboard.nixtla.io

nixtla_set_api_key(api_key="Your API key here")

2. Load sample data

df<-nixtlar::electricityhead(df)#>   unique_id                  ds     y#> 1        BE 2016-10-22 00:00:00 70.00#> 2        BE 2016-10-22 01:00:00 37.10#> 3        BE 2016-10-22 02:00:00 37.10#> 4        BE 2016-10-22 03:00:00 44.75#> 5        BE 2016-10-22 04:00:00 37.10#> 6        BE 2016-10-22 05:00:00 35.61

3. Forecast the next 8 steps ahead

nixtla_client_fcst<-nixtla_client_forecast(df, h=8, level=c(80,95))#> Frequency chosen: hhead(nixtla_client_fcst)#>   unique_id                  ds  TimeGPT TimeGPT-lo-95 TimeGPT-lo-80#> 1        BE 2016-12-31 00:00:00 45.19045      30.49691      35.50842#> 2        BE 2016-12-31 01:00:00 43.24445      28.96423      35.37463#> 3        BE 2016-12-31 02:00:00 41.95839      27.06667      35.34079#> 4        BE 2016-12-31 03:00:00 39.79649      27.96751      32.32625#> 5        BE 2016-12-31 04:00:00 39.20454      24.66072      30.99895#> 6        BE 2016-12-31 05:00:00 40.10878      23.05056      32.43504#>   TimeGPT-hi-80 TimeGPT-hi-95#> 1      54.87248      59.88399#> 2      51.11427      57.52467#> 3      48.57599      56.85011#> 4      47.26672      51.62546#> 5      47.41012      53.74836#> 6      47.78252      57.16700

Optionally, plot the results

nixtla_client_plot(df,nixtla_client_fcst, max_insample_length=200)

Anomaly Detection Using TimeGPT in 3 Easy Steps

Do anomaly detection withTimeGPT, also in 3 easy steps! Follow steps 1 and 2 from the previous section and then use thenixtla_client_detect_anomalies and thenixtla_client_plot functions.

nixtla_client_anomalies<-nixtlar::nixtla_client_detect_anomalies(df)#> Frequency chosen: hhead(nixtla_client_anomalies)#>   unique_id                  ds     y anomaly  TimeGPT TimeGPT-lo-99#> 1        BE 2016-10-27 00:00:00 52.58   FALSE 56.07623     -28.58337#> 2        BE 2016-10-27 01:00:00 44.86   FALSE 52.41973     -32.23986#> 3        BE 2016-10-27 02:00:00 42.31   FALSE 52.81474     -31.84486#> 4        BE 2016-10-27 03:00:00 39.66   FALSE 52.59026     -32.06934#> 5        BE 2016-10-27 04:00:00 38.98   FALSE 52.67297     -31.98662#> 6        BE 2016-10-27 05:00:00 42.31   FALSE 54.10659     -30.55301#>   TimeGPT-hi-99#> 1      140.7358#> 2      137.0793#> 3      137.4743#> 4      137.2499#> 5      137.3326#> 6      138.7662

nixtlar::nixtla_client_plot(df,nixtla_client_anomalies, plot_anomalies=TRUE)

Features and Capabilities

nixtlar provides access to TimeGPT’s features and capabilities, such as:

• Zero-shot Inference: TimeGPT can generate forecasts and detect anomalies straight out of the box, requiring no prior training data. This allows for immediate deployment and quick insights from any time series data.

• Fine-tuning: Enhance TimeGPT’s capabilities by fine-tuning the model on your specific datasets, enabling the model to adapt to the nuances of your unique time series data and improving performance on tailored tasks.

• Add Exogenous Variables: Incorporate additional variables that might influence your predictions to enhance forecast accuracy. (E.g. Special Dates, events or prices)

• Multiple Series Forecasting: Simultaneously forecast multiple time series data, optimizing workflows and resources.

• Custom Loss Function: Tailor the fine-tuning process with a custom loss function to meet specific performance metrics.

• Cross Validation: Implement out of the box cross-validation techniques to ensure model robustness and generalizability.

• Prediction Intervals: Provide intervals in your predictions to quantify uncertainty effectively.

• Irregular Timestamps: Handle data with irregular timestamps, accommodating non-uniform interval series without preprocessing.

Documentation

For comprehensive documentation, please refer to our vignettes, which cover a wide range of topics to help you effectively usenixtlar. The current documentation includes guides on how to:

• Get started and set up your API key
• Do anomaly detection
• Perform time series cross-validation
• Use exogenous variables
• Generate historical forecasts

The documentation is an ongoing effort, and we are working on expanding its coverage.

API Support

Are you a Python user? If yes, then check out thePython SDK forTimeGPT. You can also refer to ourAPI reference for support in other programming languages.

How to Cite

If you find TimeGPT useful for your research, please consider citing theTimeGPT-1paper. The associated reference is shown below.

Garza, A., Challu, C., & Mergenthaler-Canseco, M. (2024). TimeGPT-1. arXiv preprint arXiv:2310.03589. Available athttps://arxiv.org/abs/2310.03589

License

TimeGPT is closed source. However, this SDK is open source and available under the Apache 2.0 License, so feel free to contribute!

Get in Touch

We welcome your input and contributions to thenixtlar package!

• Report Issues: If you encounter a bug or have a suggestion to improve the package, please open anissue in GitHub.

• Contribute: You can contribute by opening apull request in our repository. Whether it is fixing a bug, adding a new feature, or improving the documentation, we appreciate your help in makingnixtlar better.