W hen we started to design Orange, the data mining suite which should first
of all satisfy our own appetites for fast prototyping of new machine
learning methods, we thought that scripting is all we needed. To some point,
we were right. Scripting is a powerful tool to combine existing components,
prototype your own methods, and write programs for data analysis (that stay
there so you can run them in exactly the same way them even after you have
published the paper about them).
But as the time went by, we realized that something is missing. Data analysis is
much about visualization. Oh, of course you can run a gnuplot from Python —
the scripting environment for Orange — and plot a graph or two. But this is
not we wanted. We were after interactive visualization. To plot a scatterplot,
but be able to choose the data points from it. Visualize a classification tree, but
be able to select nodes and corresponding data instances. Draw naive Bayesian
nomogram, and use it for interactive classification and probability prediction.
And above all that, wrap these visualization components with other standard
components for data analysis, preprocessing and modeling in a schema that is
easy to use yet powerful in enabling the user to visually program the data
analysis procedure of her needs.
Orange Widgets
This is what took us to the concepts of visual programming and interactive
graphical user interface components called widgets. Orange widgets are graphical
user interface wrappers around data analysis algorithms implemented in
Orange and Python.