search

AutoViML / Auto_TS

Automatically build ARIMA, SARIMAX, VAR, FB Prophet and XGBoost Models on Time Series data sets with a Single Line of Code. Created by Ram Seshadri. Collaborators welcome.
Apache License 2.0
729 stars 114 forks source link

Using 'ML' parameter for model_type crashes Jupyter Kernel #68

Closed Billeni07 closed 2 years ago

Billeni07 commented 2 years ago

I was able to run a test Auto_TS using all model_type parameters, except for 'ML'. Within a few moments of using it, the Jupyter Notebook Kernel crashes and restarts. Any recommendations? auto_ts version 0.0.43 yfinance version 0.1.66 Python version 3.8.8

Dependencies numpy 1.21.4 pandas 1.3.4 xlrd 2.0.1 scipy 1.7.1 prettytable 2.4.0 xgboost 1.1.1 dask 2021.10.0 matplotlib 3.4.3 seaborn 0.11.2 scikit-learn 1.0.1 statsmodels 0.12.2 pmdarima 1.8.4 fbprophet 0.7.1

AutoViML commented 2 years ago

Hi @Billeni07 👍 I think there are some version mismatch issues with your machine.

Please look at the setup.py file and compare the requirements to your machine versions above. Then try this update.

Please do the upgrade to latest version by: pip install auto-ts --upgrade

Please confirm that fixes the problem. Thanks Auto_ViML

Billeni07 commented 2 years ago

Hi @AutoViML ! Thanks for responding and trying it out. I ran setup.py file and then did the upgrade. Running the autots_univariate_example crashes the Jupiter notebook kernel at step 7.

model = auto_timeseries(score_type='rmse', time_interval='M', non_seasonal_pdq=None, seasonality=False, seasonal_period=12, model_type=['best'],dask_xgboost_flag=False, verbose=2) model.fit(traindata, ts_column,target)

I also used the ts_2.csv data file and refreshed files with Github desktop.

Is there something else I can try or provide more data about? Thanks. Billeni07

AutoViML commented 2 years ago

Hi @Billeni07 👍

can you please try to create a Colab notebook or Kaggle kernel and run the same notebook with the same data and show me? I am not seeing this error in my machine or in those public notebooks. Please share your link. Thanks