Linear Projection

../../_images/linear-projection.png

A linear projection method with explorative data analysis.

Signals

Inputs:

  • Data

    An input data set

  • Data Subset

    A subset of data instances

  • Projection

    Custom projection vectors

Outputs:

  • Selected Data

    A data subset that the user has manually selected in the projection.

  • Data

    Data with an additional column showing whether a point is selected. If more than one group is selected then also the group name is written instead.

  • Components

    Projection vectors

Description

This widget displays linear projections of class-labeled data. It supports various types of projections such as circular, linear discriminant analysis, principal component analysis, custom projection.

Consider, for a start, a projection of the Iris data set shown below. Notice that it is the sepal width and sepal length that already separate Iris setosa from the other two, while the petal length is the attribute best separating Iris versicolor from Iris virginica.

../../_images/linear-projection-stamped.png
  1. Axes in the projection that are displayed and other available axes.
  2. Optimize your projection by using Suggest Features. This feature scores attributes by average classification accuracy and returns the top scoring attributes with a simultaneous visualization update.
  3. Choose the type of projection.
  4. Axes inside a circle are hidden. Circle radius can be be changed using a slider.
  5. Adjust plot properties:
    • Set jittering to prevent the dots from overlapping (especially for discrete attributes).
    • Show legend displays a legend on the right. Click and drag the legend to move it.
    • Show class density colors the graph by class (see the screenshot below).
    • Label only selected points allows you to select individual data instances and label them.
  6. Select, zoom, pan and zoom to fit are the options for exploring the graph. Manual selection of data instances works as an angular/square selection tool. Double click to move the projection. Scroll in or out for zoom.
  7. If Send automatically is ticked, changes are communicated automatically. Alternatively, press Send.
  8. Save Image saves the created image to your computer in a .svg or .png format. Produce a report.

Example

The Linear Projection widget works just like other visualization widgets. Below, we connected it to the File widget to see the set projected on a 2-D plane. Then we selected the data for further analysis and connected it to the Data Table widget to see the details of the selected subset.

../../_images/LinearProjection-example.png

References

Koren Y., Carmel L. (2003). Visualization of labeled data using linear transformations. In Proceedings of IEEE Information Visualization 2003, (InfoVis‘03). Available here.

Boulesteix A.-L., Strimmer K. (2006). Partial least squares: a versatile tool for the analysis of high-dimensional genomic data. Briefings in Bioinformatics, 8(1), 32-44. Abstract here.