Detecting drifts

Avoiding drifts altogether in all our models is not possible, but we can aim to detect them early on and eliminate them. Here, we are going to practice drift detection with alibi_detect and evidently in Python.

Practicing with alibi_detect for drift detection

One of the widely-used Python libraries for drift detection that we want to practice with is alibi_detect. We will first import the necessary Python functions and classes and generate a synthetic dataset with 10 features and 10,000 samples using make_classification from scikit-learn:

import numpy as npimport pandas as pd
import lightgbm as lgb
from alibi_detect.cd import KSDrift
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.metrics import balanced_accuracy_score as bacc
# Generate synthetic data
X, y = make_classification(n_samples=10000, n_features=10,
    n_classes=2, random_state=42)

Then, we split the data...