Report Results?

Just posting another idea I have to see if people would want it. Would people enjoy a report_results feature? Where it prints out in a nice pdf/png document your confusion matrix, most_confused, your overall accuracy, and whatever else you want? (such as the lr plot, etc?, or give me other ideas)

I’m working on generating reports for my research and this seems less tedious than doing it manually, but let me know :smile:

It seems useful for applications that may require to generate reports on a regular basis with a fixed structure, like validation reports of a model in production :slight_smile: For the training and experimental phase maybe displaying the plots in jupyter notebooks is simpler.

Thanks for the input! I was thinking because in my research right now I’m focusing on classification and one thing that my boss wants is to see a slew of confusion matrix’s plus most confused, accuracy, etc. From a fixed structure standpoint, would it be something along that? :smile:

The end goal idea is if I’m running an experiment on a variety of different approaches, each option is grouped together for easier comparison.

@mnpinto what would go into those validation reports?

For validation reports I mean, for example, you have an app with a model that predict the temperature for several cities, then on regular basis, like weekly or monthly, you want to check how well the model is performing in comparison to observations and generate an automatic report.

But I think that it needs to be something very customizable, like having a “ReportResults” class that can handle a list of plots/tables to include in the final pdf or latex file.

Maybe this python -> latex library can be useful to achieve that: