Amazon Chronos T5 Small Forecasting Google Colab Notebook
Hi Everyone,
I am sharing this in case it is helpful, I created a Google Colab Notebook which enables testing of Chronos T5 small on small datasets for forecasting, all you need is a csv file with some data in to get started (or even if you don't have that I've provided some you can use).
Colab Notebook: https://github.com/smartaces/amazon-chronos-t5-sales-forecasting
Video walkthrough of Notebook: https://www.youtube.com/watch?v=jyrOmIiI2Bc&t=103s
I also created a series of inputs to make it easy to set forcast period lengths and intervals.
The notebook also allows for comparison against actual datasets (via csv upload), so you can see how accurate the forecasts are.
Basically this makes it super simple to test the awesomeness of Chronos.
All outputs are exported as csv file as well.
I am a big fan of this model, and have seen some really great results even with T5 small, hitting +90% accuracy.
All the best
J
Great work
I am wondering it it can be used for anomaly detection, like TimeGPT can. Do you know if there are examples of using chronos, or any other of the HF time series foundation models, that can perform anomaly detection?