Wouldn't it be nice if we can visualize the results of `models` instead of just looking at the table?

[vidyap-xgboost]

Since the goal of lazypredict is to run the models quickly and give the results quickly to get insights in a high level, visualizing these results of models or predictions would be much easier to gather insights and draw some conclusion rather than looking at a table.

Or do you think users can use matplotlib or other libraries easily and why to bother writing a function for it?

Any views regarding this?

[shankarpandala]

@vidyap-xgboost Reason why i am not plotting is, we are comparing algorithms across multiple metrics. A bar chart is space consuming and a line plot is a wrong way to plot and leads to incorrect interpretations.

If we can think of a way to represent all the important information in the table by a plot we can do it.

interactive plots using plotly etc might help. But, I want the code to run from command line not from notebook only.

So, we can give it as an option but not default.

[vidyap-xgboost]

I understand where you are coming from.. Interactive plots like plotly would be difficult to achieve from command line, don’t you think?

Giving it as an option sounds good enough. What are your thoughts around generating a simple plot by outputting as a HTML report?

[shankarpandala]

I am thinking to isolate the plotting functions, then it wouldn't matter if they are in notebook or HTML page. As this is isolated, anyone using command line won't use it at the first place.

[vidyap-xgboost]

That sounds great.

[brendalf]

I was thinking about this, one option could be a simple bar plot for only one metric that the user can choose. What you guys think? It's a faster way to release a mvp for this feature.