This is code for comparison of automatic machine learning libraries:

• auto-sklearn
• autoML from h2o
• mljar

To download datasets you need to register on openML and set OPENML_KEY in your environment.

1. Each dataset was divided into train and test set (70%/30%).
2. The autoML package was trained on train set. There was 1 hour limit for training.
3. Final autoML model was used to compute predictions on test set (samples not used for training).
4. The logloss was used to asses model performance (the lower the better).
5. The process was repeated 10 times (with different seeds), results are average over 10 repeats.