Add confusion matrix evaluation helpers for classification problems

[atecon]

Feature Request

Add confusion matrix evaluation helpers for classification problems

It would be useful for users working with classification problems to have convenient helpers in the package for evaluating the results with confusion matrices.

I propose two template functions as a starting point:

function matrix ConfusionMatrix (const series y, const series f)
    /* Confusion matrix */

    list L = y f
    xtab L --column
    matrix ret = $result[-end,-end]
    rnameset(ret, "y=1 y=2 y=3 y=4 y=5")
    cnameset(ret, "f=1 f=2 f=3 f=4 f=5")

    return ret
end function

function void PlotConfusionHeatmap (const matrix cm,
                                    const string title[null],
                                    const string filename[null])
    /* Plot a user-supplied confusion matrix with fixed heatmap settings.*/

    bundle PlotParams = _(xlabel = "Predicted class",
                ylabel = "Actual class",
                zerowhite = TRUE,
                printvals = TRUE,
                do_labels = TRUE)

    if exists(title)
        PlotParams.title = title
    endif
    if exists(filename)
        PlotParams.dest = filename
    endif

    heatmap(cm, PlotParams)
end function

These functions may serve as templates or convenience tools:

• ConfusionMatrix creates a confusion matrix from observed and predicted series.
• PlotConfusionHeatmap plots a confusion matrix with fixed heatmap settings.

Having helpers or templates like these would benefit users working on classification evaluation workflows.