Error to train using autofit

It’s my first time creating an issue so I’m sorry if I’m wrong about something 😃

I have problems when using the autofit function, when executing these lines

boosted_model = tb.ThymeBoost(verbose=0)
output = boosted_model.autofit(al_train['value'], seasonal_period=288)
predicted_output = boosted_model.predict(output, len(al_test))
tb_mae = np.mean(np.abs(al_test - predicted_output['predictions']))

I get the following error:

---------------------------------------------------------------------------
NameError                                 Traceback (most recent call last)
[<ipython-input-10-c3af4c216539>](https://localhost:8080/#) in <module>()
      1 boosted_model = tb.ThymeBoost(verbose=0)
----> 2 output = boosted_model.autofit(al_train['value'], seasonal_period=288)
      3 predicted_output = boosted_model.predict(output, len(al_test))
      4 tb_mae = np.mean(np.abs(al_test - predicted_output['predictions']))
      5 tb_rmse = (np.mean((al_test - predicted_output['predictions'])**2))**.5

[/usr/local/lib/python3.7/dist-packages/ThymeBoost/ThymeBoost.py](https://localhost:8080/#) in autofit(self, time_series, seasonal_period, optimization_type, optimization_strategy, optimization_steps, lag, optimization_metric, test_set, verbose)
    578             seasonal_sample_weights = []
    579             weight = 1
--> 580             for i in range(len(y)):
    581                 if (i) % max_seasonal_pulse == 0:
    582                     weight += 1

NameError: name 'y' is not defined

In the previous version (0.1.10) if I could execute these lines

environment:

al_train:

time	value
2022-05-28 00:00:00	3108860000.0
2022-05-28 00:05:00	3406160000.0
2022-05-28 00:10:00	3535540000.0
2022-05-28 00:15:00	3544810000.0
2022-05-28 00:20:00	3336570000.0
2022-05-28 00:25:00	2994020000.0
2022-05-28 00:30:00	3130380000.0
2022-05-28 00:35:00	2953710000.0

[…]

INFO al_train:

shape:  (840, 1)
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 840 entries, 2022-05-28 00:00:00 to 2022-05-30 21:55:00
Data columns (total 1 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   value   840 non-null    float64
dtypes: float64(1)
memory usage: 13.1 KB

[manuelchacon07]

It’s works, thanks for help me and upload the version 😃

[tblume1992]

Awesome, just remember when using autofit that the default params are:

optimization_strategy='cv', optimization_steps=3, lag=2

Which does 3 rounds of forecasts and tests on a rolling 2 period test set. Seeing that your seasonal period is 288, you may want to increase the lag (i.e. how much to test on). The total size of the holdout is optimization_steps * lag so by default 6.

If you passed:

optimization_strategy='holdout', optimization_steps=3, lag=144

Now it simply tests on the last 144 data points once (optimization_steps are ignored when strategy is holdout). May be a better strategy for autofitting with a large dataset.

Hopefully that makes sense!