Incorrect Binary Categorical Preprocessor work
Created by: andreygetmanov
- Incorrectly defines categorical features from test_pipeline_for_side_task_predict (categorical_ids = [1, 2, 3, 4, 5, 7, 8], true cat_ids = [1, 2, 3, 4, 5, 6, 8]
- column_uniques contains 1.0 and 1 as distinguished classes, so the condition of binary decoding doesn't work correctly (column is recognized as binary, if column_uniques <= 3: two classes and possible gaps)
- Feature types are falsely identified (feature_types = ["int", "str", "str", "str", "str", "str", "str", "str", "str"], true feature_types = ["int", "int", "int", "int", "str", "str", "int", "str", "str"]
Due to unstable work, the results can be unpredictable, so it's needed to understand the reason of problems, fix them and cover them by unit tests