library

In scikit-assist an experiment is defined by a dataset and a validation strategy. A dataset must be provided by a pandas DataFrame. The validation strategy is specified by a list of lists, each containing up to three indices into the dataset (training, testing and validation set).

Optionally, a list of columns that are to be used as features during model training can be provided. If not provided, all columns but the target will be used as features.

Example:

A new experiment with a DataFrame data can be specified as follows:

from skassist import Library

cross_val = [
        [tr1,te1,va1],  # 1. validation split
        [tr2,te2,va2],  # 2. validation split
        [tr3,te3,va3]   # 3. validation split
]

features = ['col1', 'col2', 'col3']

lib = Library()

lib.add('name_of_experiment', data, cross_val, features, 
        'Some meaningful desription.')

lib.exp[0].add()