Stratified Split Calculator

Calculate class-balanced train and test set sizes using stratified sampling.

About Stratified Split Calculator

Stratified sampling applies the same train percentage independently to each class, so class_n_train = class_n_total × train%, ensuring every class contributes proportionally to both the train and test sets. This calculator accepts an arbitrary list of per-class totals and produces an exact per-class train/test split, which is the standard approach for classification tasks with multiple classes of varying size.

Formula

class_train_n = round(class_total_n × train_percent / 100)

class_total_n: Total samples for class n
train_percent: Target training set percentage

Worked example

With the default values below, Stratified Split Calculator returns Total Train Samples of 8,000. Change any input above and every figure here is recalculated in your browser.

Inputs used

Class Counts (comma-separated): 5000, 3000, 1500, 500
Train %accepts 1–99 %: 80 %

Results

Total Train Samples: 8,000
Total Test Samples: 2,000
Total Samples: 10,000

Step by step

1Total samples across classes= 10000
2Total train samples= 8000

Assumptions and limits

The result is derived only from the 2 inputs listed above. Anything not among them is not part of the calculation.
In particular, this calculator does not take inflation, tax, fees or charges or compounding frequency into account.
Figures are rounded for display to whole numbers, so totals built from several rounded parts can differ slightly from the unrounded arithmetic.

How these calculators are built and tested is described on the Methodology page. If a figure looks wrong, please report it.

Frequently Asked Questions

Why split per-class instead of on the whole dataset at once?

Splitting the whole dataset randomly can under-represent rare classes in one partition purely by chance; splitting within each class guarantees every class is represented proportionally in both sets.

What if a class has very few samples?

For classes with fewer than ~10 samples, rounding can distort the train percentage significantly — consider merging rare classes or using leave-one-out validation instead.

Can I use this for train/val/test (three-way) splits?

Run this calculator twice: first split into train vs. (val+test), then split the (val+test) remainder again into val and test using the same method.