Min-Max Scaling - Machine Learning Basics

Scale a feature list to [0,1] via (x − min)/(max − min). Uses min() and max() as primitives to isolate the scaling formula; a single loop applies the transform. Library: sklearn.preprocessing.MinMaxScaler().fit_transform(X) on a column-shaped 2D array; .ravel().tolist() extracts a flat list. RESULT: scaled list (rounded).

By hand

With scikit-learn

MinMaxScaler().fit_transform(X) returns a 2D array; .ravel().tolist() flattens it. The snapshot shows the learned min and max for verification.

naive.py

data = [10, 20, 30, 40, 50]
mn = min(data)
mx = max(data)
rng = mx - mn
scaled = []
for v in data:
    scaled.append(round((v - mn) / rng, 4))
print('RESULT:', scaled)

library.py

import numpy as np
from sklearn.preprocessing import MinMaxScaler
from dalib.display import set_display
set_display()

data = [10, 20, 30, 40, 50]
X = np.array(data).reshape(-1, 1)
scaler = MinMaxScaler()
scaled = [round(v, 4) for v in scaler.fit_transform(X).ravel().tolist()]
print('min:', float(scaler.data_min_[0]))
print('max:', float(scaler.data_max_[0]))
print('RESULT:', scaled)

min: 10.0
max: 50.0
RESULT: [0.0, 0.25, 0.5, 0.75, 1.0]

Implementation notes

MinMaxScaler applies exactly (x − min)/(max − min), matching the naive formula precisely.
Sensitive to outliers: a single extreme value shifts min or max, compressing all other scaled values toward the center.
reshape(-1, 1) is required because sklearn expects a 2D feature matrix (n_samples × n_features); a 1D list is ambiguous.
Cross-reference: standardize-features (this chapter) for the mean/std alternative that is less sensitive to outliers.