TAPeri edited this page Jun 16, 2015 · 5 revisions

The Preference Learning Toolbox provides a collection of algorithms to create predictive models from ordinal data.

Is my data suited for preference learning?

Preference learning should be used when the dataset to train the model contains information about the output in an ordinal format, typically ranks, ratings or preferences, e.g.:

Data Model
Users report their top 3 favourite movies Mapping between movie cast into user preference
Users report their level of pain using a graphical pain scale Relation between symptomps and pain levels
Users are presented with pictures in pairs and select the one that shows the most frustrated person Predictor of frustration from the pixels of the images

Ranks, ratings and preferences are often stored as numbers, why shouldn't I use standard regression or classification algorithms?

When ordinal variables are treated as numerical, an important assumption is made: the distance between consecutive values is constant. When the ordinal variables are based on subjective notions, this assumption rarely (if ever) holds true. For more details, check out this paper.

How do I get started?

Prepare your dataset in the right format, download and launch the tool, and follow the steps prompted on the screen. The tool is free for scientific use. If you use PLT in your scientific work, please cite as:

 Farrugia, Vincent E., Héctor P. Martínez, and Georgios N. Yannakakis. 
 "The Preference Learning Toolbox." arXiv preprint arXiv:1506.01709 (2015)
You can’t perform that action at this time.
You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session.
Press h to open a hovercard with more details.