Decision trees are commonly used in machine learning, especially in classification. Compared to other types of models, such as neural networks, they offer a rather intuitive interpretation and they are transparent in application. Visual analytics can be used to build and refine machine learning models. For decision trees, a visual interface based on parallel coordinate plots may offer a suitable approach for analysts to evaluate and improve their models.
In this project, you plan to build an interactive visualization that enables the construction and evaluation of a particular class of decision trees, namely Fast-and-Frugal Trees. Fast-and-Frugal Trees pose some additional constraints on the tree structure, which yo can exploit in the design of the visual interface. The goal is to allow a closer integration of the human analyst in the development and evaluation of the classifiers. In particular, the visual interface will incorporate a diverse set of interactions, which are relevant at the different steps in model evaluation and refinement.
- Jianchao Han and Nick Cercone (2000) RuleViz: a model for visualizing knowledge discovery process}, In: Proc. Int. Conf. Knowledge Discovery and Data Mining, doi:10.1145/347090.347139
- Stef van den Elzen and Jarke J. van Wijk (2011) BaobabView: Interactive Construction and Analysis of Decision Trees, In: Proc. Conf. Visual Analytics Science and Technology, doi:10.1109/VAST.2011.6102453
- Gary K. L. Tam, Vivek Kothari and Min Chen (2017) An Analysis of Machine- and Human-Analytics in Classification, In: Trans. Visualization and Computer Graphics 23(1) pp. 71–80, doi:10.1109/TVCG.2016.2598829
- Nathaniel D. Phillips, Hansörg Neth, Jan K. Woike, and Wolfgang Gaissmaier (2017) FFTrees: A toolbox to create, visualize, and evaluate fast-and-frugal decision trees, In: Judgment and Decision Making 12(4) pp. 344–368