How to display categories for different models I want to visualize several different models' performance over different categories. My chart does not look very pretty so far. Additionally, it is not very easy to read the results.

How can I improve this chart or split up the chart in several ones to make it more friendly to the one it is presented to?
 A: A good rule of thumb is to think of your graphic as a standalone object. Knowing nothing about these data, I should be able to look at this graphic and have a good idea what the data analysis is about and why the graphic needed to be shown. I do not get this from the graphic in its current state, mostly because of its lack of informativeness.
A few suggestions:


*

*Truncate the Y-axis to a practical minimum, not 0%. It's a little surprising to see that the poorest performance is at 40%, but perhaps about there is where the Y-axis could start. Maybe category 2 is a fluke since it's 0 for all models? Maybe a footnote is better than wasting real-estate on a figure to have totally blank columns.

*Increase the width of the graphic.The aspect ratio for this graphic looks to be the sort of 16:9 or 5:4 which is good for a 1 1/2 column display. Create a 2 column display with a 2:1 aspect ratio. Or else split this graphic into two.

*Do not use such jarring colors. This can in fact make it harder to see the element-wise levels in a barplot. A three-tone gradient from RColorBrewer is typically easier.

*Add a legend or a caption to describe what the colors represent.

*Consider putting an asterisk on columns that are different. Use subjective or objective rules.

*Label the "categories" and the axis labels. "Category 1" means absolutely nothing to me. Use text instead of axis with an option srt=45 to write long descriptions on a bias.

*Remove the axis ticks and increase the width between groups.

*Label accuracy as a percentage instead of a decimal.

*If the categories do not have a prespecified sort-order, sort them by descending average accuracy, so one can draw their eye easily towards categories that present the most challenging validations.
