add special dates to 02_holiday

[khuyentran1401]

No description provided.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[elessar-for-github]

Changes made by this pull request:

[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	0.7329	0.5385	0.0043	0.0033

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	0.4993	0.5608	0.0037	0.0035

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	0.4772	0.5966	0.0046	0.0044

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	0.5848	0.5116	0.005	0.0045

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	0.6848	0.6481	0.0051	0.0046

Plot:

[Copilot]

Copilot wasn't able to review any files in this pull request.

[ngupta23]

Line #7.            start=df.iloc[0][time_col], end=df.iloc[-1][time_col], freq="D"

This assumes data is sorted which it might not be. maybe use min() and max() if the time column (assuming it is datetime format).

Also, just take the frequency as an argument and use it here. This will make the function generic so users can use it for other frequencies also. If you do this, rename the variable monthly_features to just features

Reply via ReviewNB

[khuyentran1401]

Great suggestion!

[khuyentran1401]

@ngupta23 I noticed that the original tutorial uses "D" for dates_range but "M" for resample, so should I keep freq="D"?

def add_date_features_to_dataframe(df, date_extractor, time_col="month", freq="M"):
    # Create a copy of the dataframe
    df = df.copy()

    # Create date range
    dates_range = pd.date_range(
        start=df[time_col].min(), end=df[time_col].max(), freq="D"
    )

    # Get date feature indicators and resample to specified frequency
    features_df = date_extractor(dates_range)
    features = features_df.resample(freq).max()
    features = features.reset_index(names=time_col)

    # Merge with input dataframe
    result_df = df.merge(features)

    return result_df

[ngupta23]

Ah ok. I think it kept D for date range since the holidays will not be perfectly aligned with W and M freq (could occur on any day of the week/month. I guess your changes will work with resampling to freq, but keep date_range to use D.

[khuyentran1401]

sounds good

[ngupta23]

change "current" to "historical"?

Reply via ReviewNB

[ngupta23]

Line #1.    fcst_df_holidays = nixtla_client.forecast(

Change to long horizon model to remove warning?

Reply via ReviewNB

[ngupta23]

Line #4.        freq="M",

Change M to ME to remove the warning from the output.

Reply via ReviewNB

[ngupta23]

Line #1.    nixtla_client.weights_x.plot.barh(x='features', y='weights', figsize=(10, 10))

At some point we had discussed to streamline this and not use the weights_x attribute, instead standardizing on the use of shap which is also what is shown in the interpretibility tutorial. Can you check with @marcopeix and @elephaint if we want to continue to show this?

Reply via ReviewNB

[ngupta23]

remove "also"

Reply via ReviewNB

[ngupta23]

Line #4.        freq="M",

Same comments as above - change to ME and maybe use the long-horizon model?

Reply via ReviewNB

[ngupta23]

Line #1.    nixtla_client.weights_x.plot.barh(x="features", y="weights", figsize=(10, 6))

Same comment as above - maybe use Shap?

Reply via ReviewNB