This example uses Scikit-Learn's make_classification method to create 2-class datasets that begin to become indistinguishable over many trials. In the image above, note that the score, which is the mean accuracy, decreases with each trial.

In params.yaml, both the dataset and the classifier parameters are set. The dataset value "class_sep" determines the initial class separation. This value decreases with each trial.

The model implementations and the plotting functionality are in clf_test.py.

Clone this repository.

~$ cd model_drift_testing

Activate your Python Machine Learning Environment

~$ python main.py

The program will end when the score falls below the threshold set in main.py