Mastering Feature Binning with sklearn: Uniform, Quantile, and K‑Means Methods

When building classification models, feature engineering often requires discretizing continuous variables into categorical bins, which can make models more stable and reduce over‑fitting.

Equal‑width binning

Equal‑frequency (quantile) binning

Clustering (k‑means) binning

The following example uses a synthetic score dataset to illustrate the concepts.

import numpy as np
import pandas as pd
np.random.seed(1)

n = 20
ID = np.arange(1, n+1)
SCORE = np.random.normal(80, 10, n).astype('int')
df = pd.DataFrame({'ID': ID, 'SCORE': SCORE})

scikit‑learn's KBinsDiscretizer class performs the binning operations.

Parameter n_bins: number of bins (default 5).

Parameter strategy: binning strategy, options are:

uniform – equal‑width bins

quantile – equal‑frequency bins

kmeans – clustering‑based bins

Parameter encode: whether to encode the binned feature (e.g., ordinal, one‑hot).

Since KBinsDiscretizer expects a column vector, reshape the data:

score = df['SCORE'].values.reshape(-1, 1)

Equal‑width binning

Divide the data into three bins.

from sklearn.preprocessing import KBinsDiscretizer

dis = KBinsDiscretizer(n_bins=3,
                      encode="ordinal",
                      strategy="uniform")
label_uniform = dis.fit_transform(score)  # transformer

With a minimum value of 56 and a maximum of 97, the bin edges are [56, 69.6667, 83.3333, 97].

Equal‑frequency binning

Each bin contains roughly the same number of samples. Using three bins:

dis = KBinsDiscretizer(n_bins=3,
                      encode="ordinal",
                      strategy="quantile")
label_quantile = dis.fit_transform(score)

Clustering binning

First cluster the continuous variable, then use the cluster label as the bin identifier.

dis = KBinsDiscretizer(n_bins=3,
                      encode="ordinal",
                      strategy="kmeans")
label_kmeans = dis.transform(score)  # transformer

Comparison

df["label_uniform"] = label_uniform
df["label_quantile"] = label_quantile
df["label_kmeans"] = label_kmeans

References:

https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.KBinsDiscretizer.html

https://mp.weixin.qq.com/s/H19asF7Qo_0Wc5FIn8Qkww