ml_configs

This module contains configurations for the mltrainer.py module.

It includes the following configurations:

• Maximum number of days to consider in the models (MAX_N_DAYS)

• Test size for splitting the data into training and testing sets (TEST_SIZE)

• Filters for the tags used in the models (TAG_FILTERS)

• Additional tags to be added to the models’ features (TAGS_TO_ADD)

• Tags to be dropped from any dataset that contains them (TAGS_TO_DROP)

Each tag filter is a dictionary that specifies the minimum and maximum values (or percentiles) for the tag. If the ‘replace’ key is present and set to True, values outside the min/max range are replaced with the result of applying an aggregation method to the values within the min/max boundary. Otherwise, values outside the min/max range are discarded.

The additional tags are added to the models’ features without interfering with the original order of the columns. This is necessary because the target column from each model is represented by the last column of each dataset, and the source code relies on this fact to work properly.

The tags to be dropped are removed from any dataset that contains them.