Frequency Polygon Results
The graph, frequency table, statistics, and ML diagnostics update after calculation.
Frequency Distribution Table
Dataset Statistics
Dataset Comparison and Distribution Shift
Exportable Result Summary
Export and Sharing
Formula Used
How to Use
- Select raw values or grouped class intervals.
- Enter one or more datasets, or upload a CSV file.
- Choose an automatic bin rule or custom class settings.
- Select the displayed frequency type and preprocessing options.
- Customize the chart and distribution diagnostics.
- Calculate, inspect warnings, then export the graph and tables.
Example Data Table
| Dataset | Example values | ML interpretation |
|---|---|---|
| Training confidence | 0.42, 0.51, 0.63, 0.71, 0.82, 0.93 | Model confidence distribution during training. |
| Validation confidence | 0.31, 0.44, 0.56, 0.64, 0.72, 0.84 | Comparison with unseen validation observations. |
| Anomaly scores | 0.05, 0.08, 0.11, 0.14, 0.60, 0.91 | Sparse high-score intervals may indicate anomalies. |
Frequency Polygons in Machine Learning
A frequency polygon connects class midpoint frequencies with straight segments. It makes distribution shapes easier to compare. Multiple datasets remain readable on one chart.
Machine learning teams use polygons for feature inspection and drift checks. Training and testing distributions can be compared quickly. Unexpected gaps may reveal sampling problems.
A histogram emphasizes filled class areas, while polygons emphasize shape. Frequency polygons also support overlapping datasets clearly. This makes model diagnostics more compact.
Frequently Asked Questions
What is a frequency polygon?
It is a line graph joining frequencies plotted at class midpoints.
Why are zero-frequency endpoints useful?
They visually close the polygon at both ends of the distribution.
Which bin rule should I choose?
Sturges suits general data, while Freedman–Diaconis handles skew and outliers better.
Can unequal class widths be used?
Yes, but frequency density usually provides a fairer comparison.
Can I compare training and testing data?
Yes. Add both datasets and use shared intervals for direct comparison.
What does PSI indicate?
Population Stability Index measures distribution change between a reference and comparison dataset.
What does Jensen–Shannon divergence indicate?
It measures how different two probability distributions are using a symmetric score.
How are invalid values handled?
They can be ignored automatically or treated as calculation-stopping errors.
Can CSV columns become separate datasets?
Yes. Each numeric CSV column is imported as a named dataset.