Boosting Feature Importance Plot Calculator

Calculated Results

Interpretation

Stability and Agreement

Feature Ranking Table

Rank	Feature	Model	Method	Group	Raw	Mean	Median	Std. Dev.	Minimum	Maximum	Normalized	Percent	Cumulative	Runs	Status

Data and Model Inputs

Example preset

Upload CSV, TSV, or JSON

Text delimiter

Feature importance data

Required columns are feature and importance. Model, method, fold, and group are optional.

Processing Options

Importance method label

Repeated-row aggregation

Comparison series

Scaling

Sorting

Feature grouping

Prefix separator

Model filter

Method filter

Feature-name filter

Exclude features

Highlight features

Display top N

Use zero to include every feature.

Also include bottom N

Minimum absolute importance

Highlight threshold

Cumulative target (%)

Decimal places

Hide zero importance

Rank by absolute value

Show error bars

Combine duplicate rows

Chart Options

Plot type

Chart title

Chart subtitle

Value-axis label

Category-axis label

Legend position

Chart height 620 px

Bar spacing 0.25

Marker size 10

Chart font size 13

Main bar colour

Highlight colour

Show gridlines

Logarithmic value axis

Show value labels

Wrap long feature names

Formula Used

Normalized Importance_i = I_i ÷ Σ |I_j|

Importance Percentage_i = Normalized Importance_i × 100

Cumulative Importance_k = Σ_i=1...k |Normalized Importance_i|

Mean Importance_i = Σ I_ir ÷ m

Standard Deviation_i = √[Σ(I_ir − Ī_i)² ÷ (m − 1)]

Importance definitions differ across boosting libraries. Compare only compatible methods and scales. Treat importance as model evidence, not causal proof.

How to Use

Paste feature importance data or upload a supported file.
Select aggregation, scaling, filtering, and comparison options.
Choose a chart style and presentation settings.
Press Calculate and Plot to generate ranked results.
Review warnings, cumulative coverage, and stability information.
Export the table, chart, or complete PDF report.

Example Data Format

feature	importance	model	method	fold	group
annual_income	0.286	XGBoost	Gain	1	Financial
credit_score	0.244	XGBoost	Gain	1	Financial
account_age	0.151	XGBoost	Gain	1	History

Frequently Asked Questions

What is boosting feature importance?

It estimates each predictor's contribution inside a fitted boosting model.

How does gain importance work?

Gain measures average improvement produced by splits using each feature.

What is split importance?

Split importance counts how frequently trees use a given feature.

What is cover importance?

Cover reflects observations affected by splits involving each feature.

Why do libraries produce different results?

Libraries use different definitions, defaults, scales, and tree-building procedures.

Should importance values total 100 percent?

Only normalized values total 100 percent. Raw values usually do not.

Can feature importance values be negative?

Permutation importance can be negative when shuffling improves model performance.

How do correlated features affect importance?

Correlated predictors may divide, mask, or redistribute measured importance unfairly.

What is permutation importance?

It measures performance loss after randomly shuffling one feature's values.

How is SHAP importance different?

SHAP summarizes local contribution magnitudes using a game-theoretic attribution method.

Does feature importance prove causation?

No. Importance describes model behavior and never establishes causal relationships.