What type of visualization is most appropriate for stakeholders to compare model performance across multiple categories?

The most appropriate visualization for stakeholders to compare model performance across multiple categories is a bar chart with error bars for each category. This type of chart clearly presents categorical data where the height of each bar represents the performance measure for each category. The addition of error bars provides valuable context by illustrating the variability or uncertainty around the performance estimates, allowing stakeholders to easily assess both the average performance and its reliability.

In situations where comparisons between distinct categories are essential, the clarity and straightforwardness of a bar chart make it easy for stakeholders to grasp differences at a glance. The error bars enhance this by indicating how much confidence stakeholders can have in the measurements, which is critical when making decisions based on model performance. This visual representation effectively conveys both the central tendency and variability across categories, delivering a comprehensive view of performance.

Other types of visualizations serve different purposes. For instance, a line chart is typically used to illustrate trends over time rather than comparing discrete categories. A scatter plot is better suited for showing relationships between two continuous variables, not for comparing multiple categories. A pie chart focuses on the proportion of each category within a whole but does not lend itself well to performance comparisons as it does not convey quantitative differences in a straightforward manner.