Feature Distribution by Class Calculator

Compare numeric or categorical feature distributions across classes, explore overlap and separation, apply preprocessing, inspect statistics, detect outliers, and export polished reports instantly online.

Result summary

Load or paste data, map the feature and class columns, then calculate the distribution.

Waiting for data

Interactive distribution graph

Use the legend, zoom controls, hover details, and selection tools.

Data and analysis controls

Paste CSV, TSV, semicolon-separated, or whitespace-separated values. The first row should contain column names.

Include, rename, reorder, or recolor detected classes.

UseOriginal classDisplay labelColorOrder
Parse data to detect classes.

Class-level descriptive statistics

No statistics available.

Class separation and pairwise comparisons

No pairwise results available.

Outlier and missing-value report

No outlier report available.

Cleaned dataset preview

No cleaned data available.

No calculation has been completed.

Formula used

Mean: x̄ = Σx / n

Sample variance: s² = Σ(x − x̄)² / (n − 1)

Z-score: z = (x − x̄) / s

Cohen’s d: d = (x̄₁ − x̄₂) / spooled

Fisher score: between-class variance divided by within-class variance.

KDE: f̂(x) = [1 / (nh)] ΣK((x − xᵢ) / h)

Overlap coefficient: the numerical integral of the smaller class density.

How to use

  1. Paste data or upload a delimited file.
  2. Parse the dataset and map each required column.
  3. Select a distribution graph and preprocessing method.
  4. Choose visible classes and adjust their display settings.
  5. Calculate, inspect results, then export the required report.

Example data

sepal_lengthspeciesweightlabel
5.1Setosa1Flower 1
6.4Versicolor1Flower 2
7.1Virginica1Flower 3

Frequently asked questions

What does this graph show?

It shows how one feature varies across target classes.

Which chart should I choose?

Use histograms for counts and KDE curves for shape.

Can categorical features be compared?

Yes. Choose categorical mode and frequency bars.

Why are small classes risky?

Small samples create unstable densities and misleading summaries.

What indicates useful class separation?

Large effect sizes and low overlap suggest stronger separation.

Does statistical significance guarantee usefulness?

No. Practical effect size and validation still matter.

How are missing values handled?

They can be removed or imputed within classes.

Can outliers be excluded?

Yes. Select a method and choose exclusion.

Why use shared histogram bins?

Shared bins make class comparisons more direct and fair.

What does Fisher score measure?

It compares between-class separation against within-class spread.

Can I export the analysis?

Yes. Export charts, tables, cleaned data, and reports.