miur-daad-balancing 1.0.0

Testing dataset balancing techniques from previous works.

Testing dataset balancing techniques from previous works.

How do I install this package?

As usual, just download it using pip:

pip install miur_daad_balancing

Tests Coverage

Since some software handling coverages sometime get slightly different results, here’s three of them:

Usage

Three balancing methods are available for the MIUR-DAAD project:

Umbalanced

This method just leaves the data as-is, and is used more as callback usefull to uniform the pipeline:

from miur_daad_balancing import umbalanced

training, testing = generate_my_data(...)
balanced_training, balanced_testing = umbalanced(training, testing)

Balanced

Applies a maximum threshold to every class in the training set as specified in the default package settings (3000):

from miur_daad_balancing import balanced

training, testing = generate_my_data(...)
balanced_training, balanced_testing = balanced(training, testing)

Full Balanced

Applies a maximum threshold to every class in the training set and balances to some default proportions the testing set:

from miur_daad_balancing import full_balanced

training, testing = generate_my_data(...)
balanced_training, balanced_testing = full_balanced(training, testing)