Out-of-Bag Error Curve Calculator

Calculated result

Results update after calculation and remain printable.

Classification

Recommended number of trees

—

Enter or load data, then calculate the curve.

Selected model

—

Minimum OOB metric

—

Final OOB metric

—

Stabilization point

—

Initial value

—

Absolute improvement

—

Relative improvement

—

OOB accuracy at recommendation

—

Model and analysis settings

Task type

Primary OOB metric

Tree selection method

Selected model

Confidence level

Decimal precision

Total observations

Number of features

Features per split

Random seed

Bootstrap sample size

Minimum samples per leaf

Maximum tree depth

Use zero for unlimited depth.

Positive class

Smoothing and stabilization controls

Smoothing method

Smoothing window

Exponential strength

Current value: 0.35

Stability tolerance

Consecutive stable points

Percentage-change threshold

OOB weight in combined score

Current value: 0.65

Show raw curve

Show smoothed curve

Show confidence band

Show training error

Show validation error

Show class-specific curves

Data input

Use proportions such as 0.152 for 15.2%, unless a regression metric uses original units.

Paste CSV or tab-separated data

Upload CSV file

Recognized columns match the table headings. Unknown columns are ignored.

#	Model	Trees	OOB observations	Incorrect	Correct	OOB metric	Training error	Validation error	Std. deviation	Class errors	Actions

Chart customization

Chart title

Horizontal-axis label

Vertical-axis label

Line thickness

Marker size

Line style

Legend position

Vertical scale

Y-axis minimum

Y-axis maximum

Chart height

Show grid

Show legend

Show tooltips

Crosshair cursor

Interactive OOB curve

Zoom, pan, inspect points, hide curves, and reset axes using the Plotly toolbar.

Detailed statistics

No calculated statistics yet.

Export and reporting

No results yet.

Formula used

Classification OOB error: Incorrect OOB predictions ÷ total OOB predictions

OOB accuracy: 1 − OOB error rate

Regression MSE: Σ(actual − predicted)² ÷ n

Regression RMSE: √MSE

Regression MAE: Σ|actual − predicted| ÷ n

Relative improvement: (initial metric − final metric) ÷ initial metric × 100%

Confidence interval: metric ± z × standard error

Moving average: mean of values inside the selected rolling window

How to use this calculator

Select classification or regression mode.
Choose the primary metric and selection rule.
Enter tree checkpoints and OOB results.
Adjust smoothing, confidence, and stability settings.
Calculate the curve and inspect diagnostics.
Export data, charts, or a PDF report.

Frequently asked questions

What is an out-of-bag observation?

It is a training observation omitted from one bootstrap sample. That tree can evaluate the omitted observation independently.

Why does OOB error estimate generalization?

Each prediction uses trees that did not train on that observation. This approximates repeated validation without a separate split.

How is the recommended tree count selected?

The default rule finds the first stable point. Other rules can select minimum error, validation error, or one-standard-error results.

Can OOB error increase after more trees?

Small increases are normal because finite samples create noise. Persistent increases may signal data drift, instability, or configuration problems.

What does the one-standard-error rule do?

It selects the earliest simpler checkpoint within one standard error of the best observed metric.

Can I compare multiple forest models?

Yes. Give rows different model names, then select or display each model curve.

How should percentages be entered?

Enter rates as proportions. For example, enter 0.15 to represent fifteen percent.

Does regression mode support original metric units?

Yes. MSE, RMSE, and MAE values can use their natural scales.

When are more trees unnecessary?

More trees add little value after the curve stabilizes. The plateau diagnostics identify that region.