I am involved in a project that adds value in visualizing misclassification in the text mining domains. I am wondering whether anyone has experience in formally proofing that the visualisations are in fact aiding the overall data science project outcome.
There is no way to "prove" that a method is helpful.
You have to present results in a consistent way, show examples and make compelling arguments
Let say for "my student", I would recommend to compare and try out alternative approaches: 1) for subjective evaluation we can rely on human imagination Article The VVIQ and beyond: Vividness and its measurement.
2) for objective evaluation = use ITR (Information Transfer Rate) because visualization objectively is encoding the same information by filtering out less significant details, but subjective factor (human imagination) is important to turn on imagination for decoding and interpretation of visual information presented.
It all depends on the task. Usually, visualization aims at understanding a phenomenon. Look at "ISO 9241-210:2019 Ergonomics of human-system interaction — Part 210: Human-centred design for interactive systems" for the user-centered process. Concretely:
(1) define what the user is supposed to understand in terms of criteria;
(2) design your visualization;
(3) evaluate the understanding through usability testing.
This process allows to compare the performance of different visualizations. It usually iterates between steps 2 and 3.
It all depends on the task. Usually, visualization aims at understanding a phenomenon. Look at "ISO 9241-210:2019 Ergonomics of human-system interaction — Part 210: Human-centred design for interactive systems" for the user-centered process. Concretely:
(1) define what the user is supposed to understand in terms of criteria;
(2) design your visualization;
(3) evaluate the understanding through usability testing.
This process allows to compare the performance of different visualizations. It usually iterates between steps 2 and 3.
Best practices for quantitative data visualisation:
Mainly, these two books from Stephen Few: “Show Me the Numbers” and “Now You See It”
For more info on data visualization: http://www.perceptualedge.com/
To know more about the psychology of visual perception, I suggest Colin Ware’s two books: “Visual Thinking for Design” and “Information Visualization: Perception for Design”
https://ccom.unh.edu/vislab/colin_ware.html
You might find some useful hints on the advantages of visualization by looking into research on "learning modalities." Visualization in general is useful for most people, regardless of the subject matter. I think phycologists and educators are likely to have some good research to back this up... and their research might give you some good ideas for application in data science.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining