docs: add tutorial on custom frequencies

[MMenchero]

Description

This tutorial shows how to pass a custom frequency to TimeGPT, in particular, the days the US stock market is open.
This is necessary as several of our users working on financial forecasting use this frequency.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[github-actions]

Experiment Results

Experiment 1: air-passengers

Description:

variable	experiment
h	12
season_length	12
freq	MS
level	None
n_windows	1

Results:

metric	timegpt-1	timegpt-1-long-horizon	SeasonalNaive	Naive
mae	12.6793	11.0623	47.8333	76
mape	0.027	0.0232	0.0999	0.1425
mse	213.936	199.132	2571.33	10604.2
total_time	0.7636	0.8035	0.0042	0.0032

Plot:

Experiment 2: air-passengers

Description:

variable	experiment
h	24
season_length	12
freq	MS
level	None
n_windows	1

Results:

metric	timegpt-1	timegpt-1-long-horizon	SeasonalNaive	Naive
mae	58.1031	58.4587	71.25	115.25
mape	0.1257	0.1267	0.1552	0.2358
mse	4040.21	4110.79	5928.17	18859.2
total_time	1.0719	0.9171	0.0048	0.0034

Plot:

Experiment 3: electricity-multiple-series

Description:

variable	experiment
h	24
season_length	24
freq	H
level	None
n_windows	1

Results:

metric	timegpt-1	timegpt-1-long-horizon	SeasonalNaive	Naive
mae	178.293	268.13	269.23	1331.02
mape	0.0234	0.0311	0.0304	0.1692
mse	121589	219485	213677	4.68961e+06
total_time	0.8005	2.8258	0.0048	0.0043

Plot:

Experiment 4: electricity-multiple-series

Description:

variable	experiment
h	168
season_length	24
freq	H
level	None
n_windows	1

Results:

metric	timegpt-1	timegpt-1-long-horizon	SeasonalNaive	Naive
mae	465.497	346.972	398.956	1119.26
mape	0.062	0.0436	0.0512	0.1583
mse	835021	403760	656723	3.17316e+06
total_time	0.8551	0.9843	0.005	0.0044

Plot:

Experiment 5: electricity-multiple-series

Description:

variable	experiment
h	336
season_length	24
freq	H
level	None
n_windows	1

Results:

metric	timegpt-1	timegpt-1-long-horizon	SeasonalNaive	Naive
mae	558.673	459.757	602.926	1340.95
mape	0.0697	0.0565	0.0787	0.17
mse	1.22723e+06	739114	1.61572e+06	6.04619e+06
total_time	0.987	1.0628	0.005	0.0046

Plot:

[marcopeix]

I think it would be good to also include the integer frequency for highly irregular time series. Wdyt?

[MMenchero]

I thought about what you suggested @marcopeix but I don't want to add another dataset as the tutorial already has 2. If it’s possible to use the plt dataset, I’d really appreciate it if you could let me know.

[marcopeix]

I thought about what you suggested @marcopeix but I don't want to add another dataset as the tutorial already has 2. If it’s possible to use the plt dataset, I’d really appreciate it if you could let me know.

You can use the PLT dataset. The ds column would have to be integers, and then you specify freq=1. Drawback of that is that you then don't know the specific date. Maybe it would over complicate the tutorial, so never mind!

PR looks good to me! Quick reminder, you must update the mint.json file with the new tutorial notebook!

[MMenchero]

Thanks @marcopeix since the nb already exists, using the old name should be equivalent to changing the mint.json.

This is ready for review @AzulGarza

[marcopeix]

Line #4.        'https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/air_passengers.csv',

Maybe we can set the url in a separate variable and use it in the function? I think it would make the code cleaner.

Reply via ReviewNB

[MMenchero]

Done.

[marcopeix]

Line #8.    polars_df 

I think we can remove this, since it won't be displayed in the notebook

Reply via ReviewNB

[MMenchero]

Removed.

[marcopeix]

Line #1.    pltr_df = pd.read_csv('https://raw.githubusercontent.com/Nixtla/transfer-learning-time-series/main/datasets/openbb/pltr.csv', parse_dates=['date'])

Same comment as before for the url. Feel free to disregard if you don't feel it's relevant.

Reply via ReviewNB

[MMenchero]

Done, agree it is cleaner.

[marcopeix]

LGTM