NumPy Random Module

Why random matters in analytics

Random values help with:

• Simulation and Monte Carlo experiments
• Creating synthetic datasets for testing
• Sampling
• Bootstrapping

numpy.random supplements Python’s built-in random module with functions that generate whole arrays of samples at once. Python’s random module only produces one value at a time, so for large samples NumPy is dramatically faster.

Recommended API: default_rng()

Modern NumPy recommends using a Generator object instead of the old global numpy.random functions:

rng

import numpy as np
 
rng = np.random.default_rng(42)
print(rng.integers(1, 10, size=5))

diagram Generating random data with a Generator mermaid

  flowchart LR
  A["np.random.default_rng(seed)"] --> B["Generator object (rng)"]
  B --> C["rng.integers(...)"]
  B --> D["rng.normal(...)"]
  B --> E["rng.choice(...)"]
  B --> F["rng.shuffle(...)"]
  C --> G["Same seed -> same numbers every run"]
  D --> G
  E --> G

A seeded Generator object produces the same reproducible stream of pseudorandom values every run.

Seeding (reproducibility)

Seeding ensures you get the same results each run — essential for reproducible experiments and tutorials.

seed

import numpy as np
 
rng = np.random.default_rng(123)
print(rng.normal(size=3))

From the book

McKinney recommends creating one explicit Generator object (via np.random.default_rng(seed=...)) and reusing it throughout your code, rather than calling the older global numpy.random.* functions. The generator’s internal state changes every time you draw from it, and keeping it isolated in one object makes your random draws easier to reason about and reproduce.

Random integers

integers

import numpy as np
 
rng = np.random.default_rng(7)
arr = rng.integers(low=0, high=100, size=10)
print(arr)

Random floats

random

import numpy as np
 
rng = np.random.default_rng(7)
arr = rng.random(5)  # uniform in [0, 1)
print(arr)

Normal distribution

normal

import numpy as np
 
rng = np.random.default_rng(7)
arr = rng.normal(loc=0, scale=1, size=5)
print(arr)

Choice (sampling)

choice

import numpy as np
 
rng = np.random.default_rng(7)
categories = np.array(["A", "B", "C"])
print(rng.choice(categories, size=10, replace=True))

Weighted sampling:

choice-weights

import numpy as np
 
rng = np.random.default_rng(7)
values = np.array(["low", "medium", "high"])
probs = [0.6, 0.3, 0.1]
print(rng.choice(values, size=10, p=probs))

Shuffle

shuffle

import numpy as np
 
rng = np.random.default_rng(7)
arr = np.arange(10)
rng.shuffle(arr)
print(arr)

Synthetic dataset example

synthetic

import numpy as np
 
rng = np.random.default_rng(0)
 
n = 100
age = rng.integers(18, 60, size=n)
income = rng.normal(loc=60000, scale=15000, size=n)
 
X = np.column_stack([age, income])
print(X.shape)

Next

Continue to: Linear Algebra with NumPy to learn dot products, matrix multiplication, and solving systems.

🧪 Try It Yourself

Exercise 1 – A Seeded Generator

pch.exerciseTag pch.exerciseTitle

Exercise 2 – Weighted Sampling

pch.exerciseTag pch.exerciseTitle

Exercise 3 – Shuffle In Place

pch.exerciseTag pch.exerciseTitle