You can pass the output of the model to the profiling system to monitor if things are drifting.
It's also possible to monitor the input data and link back.
There's quite a few ways to do this, but effectively you can monitor drift by identifying which inputs have the greatest impact in accuracy. Then tying that back to predict the drift over time.
Wow! A great idea (haven't look into the code yet). With the new EU AI regulations coming in 2023/4. Every company with ML in production will need to be able to monitor these issues. Potential for a very good open core business model.
Effectively, you can monitor changes between profiles:
data1 = dp.Data("file_a.csv") # Load a CSV file
profile1 = dp.Profiler(data1) # Generate a profile
data2 = dp.Data("file_b.csv") # Load another CSV file
profile2 = dp.Profiler(data2) # Generate another profile
diff_report = profile1.diff(profile2)
print(json.dumps(diff_report, indent=4))
The system we have generates reports, it might be worth adding it OP.