Thank you Abrita. Now I can preproduce the plots.
Note that half of the plots are each other's Inverse (reversed axes). If we are looking for an overview of all uniquely mapped features we should have (n^2-n)/2 plots i.e. for n=4 features, this gives us 6 plots, for example:
data = ExampleData[{"MachineLearning", "FisherIris"}, "Data"];
data /. Rule[x__, #] :> x /. Rule[x__, _] :> Nothing & /@ {"setosa",
"versicolor", "virginica"};
ListPlot[%[[All, All, #]],
PlotMarkers -> {{Red, 6}, {Green, 6}, {Blue, 6}},
Frame ->
True] & /@ (Sort /@ Tuples[Range[4], 2] /. {x_, x_} :> Nothing //
Union)