Closed gabrielefantini closed 11 months ago
Hello. Thanks for your attention to the framework. It's actually problem with statsmodels and we can't easily solve it.
I can to propose this solution. It looks more complicated but actually I removed glm operation from list of models and choose assumption without glm node
operations = OperationTypesRepository(operation_type='all').suitable_operation(TaskTypesEnum.ts_forecasting).remove(
'glm')
task = Task(TaskTypesEnum.ts_forecasting, TsForecastingParams(forecast_length=2))
fedot_model = Fedot(problem='ts_forecasting', task_params=task.task_params, timeout=1,
available_operations=operations,
initial_assumption=PipelineBuilder().add_node('lagged').add_node('ridge').build())
input_data = InputData(idx=np.arange(0, 10),
features=np.array([1, 2, 3, 4, 5, 6, 7, 8, 8, 10]),
target=np.array([1, 2, 3, 4, 5, 6, 7, 8, 8, 10]),
task=task,
data_type=DataTypesEnum.ts)
train_input, validation_input = train_test_data_setup(input_data)
pipeline = fedot_model.fit(train_input)
Thanks a lot!
Partly solved (https://github.com/aimclub/FEDOT/pull/1153), partly depends on thrid-party libs.
Hi, I'm trying the auto forecasting feature. I've already tried it on complex datasets and it works great! Now I'm struggling using a dumb dataset.
Here is my code:
Why it gives me the following error? " statsmodels.tools.sm_exceptions.PerfectSeparationError: Perfect separation detected, results not available"
Thanks you in advance, Gabriele.