Andrews Curves Calculator

Calculated results

Interactive Andrews curves

Formula and generated structure

Analytical report

Class summary

Feature order and coefficients

Processed data

Curve values at selected points

Nearest observations and curve distances

Outlier analysis

1. Dataset input

Paste CSV or TSV data *

Upload CSV or TSV

Delimiter

First row contains headings

Load or paste a dataset, then select Parse data.

Data preview

No dataset has been parsed.

2. Columns and observation filters

Observation label column

Class or category column

Numeric feature columns

Visible classes

Observation label search

Specific row numbers

Leave empty to include every matching row.

3. Preprocessing

Missing-value handling

Feature transformation

Scaling method

Winsorization

Displayed decimal places

Comparison mode

4. Feature ordering

Automatic ordering method

Manual feature order

Parse data and select features first.

5. Curve domain and performance

Minimum t

Maximum t

Curve points

Maximum visible curves

Large-data sampling

Pairwise distance limit

Outlier method

Outlier threshold

6. Plot appearance and grouping

Chart title

X-axis title

Y-axis title

Individual line width

Individual opacity

Plot height

Chart theme

Legend position

Domain tick labels

Hover mode

Show individual curves

Show class means

Show class medians

Show confidence bands

Show zero reference

Show gridlines

Show legend

Highlight outliers

Formula used

Each multivariate observation becomes a Fourier curve. The first feature contributes a constant term.

The remaining features alternate between sine and cosine terms. Their frequencies increase with feature position.

f(t) = x₁/√2 + x₂ sin(t) + x₃ cos(t) + x₄ sin(2t) + x₅ cos(2t) + …

How to use

Paste a table or upload a CSV file.
Parse the dataset and confirm detected columns.
Select numeric features, labels, and an optional class.
Choose preprocessing, feature order, and curve settings.
Generate the chart and inspect separation or overlap.
Download coordinates, processed data, images, or reports.

Example data

The sample uses four flower measurements and three species. Standardization makes differently scaled measurements comparable.

Sample	Class	Sepal length	Sepal width	Petal length	Petal width
Iris-001	Setosa	5.1	3.5	1.4	0.2
Iris-051	Versicolor	7.0	3.2	4.7	1.4
Iris-101	Virginica	6.3	3.3	6.0	2.5

Frequently asked questions

What are Andrews curves?

They map every multivariate row into a continuous Fourier curve. Similar rows usually produce similar curve shapes.

Why should features be standardized?

Large-scale features can dominate the curve. Standardization gives features more comparable influence.

Does feature order matter?

Yes. Early features receive lower-frequency terms and often shape broader curve patterns.

How do curves reveal class separation?

Classes that occupy distinct curve regions may be easier to distinguish. Heavy overlap suggests similar multivariate profiles.

Can this detect outliers?

The calculator compares each curve with its class mean. Large distances are flagged using configurable thresholds.

What domain should I use?

The conventional interval is minus pi through pi. Other complete periods can also work.

How many points are needed?

Two hundred fifty points usually look smooth. Increase resolution for exports or detailed inspection.

What happens with missing values?

Rows can be removed or values can be imputed. The chosen method appears in the report.

Can large datasets be plotted?

Yes, but too many curves become cluttered. Use sampling, filtering, and class averages.

What does curve distance mean?

It summarizes shape differences across sampled t values. Smaller distances indicate more similar observations.

Can I export the analysis?

Processed data, curve coordinates, outliers, images, and a PDF report are available.