Visualizing drifting features
After ranking the univariate results, you know that drift hotel
and country
features are impacting the model's performance the most. In this exercise, you will look at the drift results and distribution plots of them to determine the root cause of the problem.
The results from the univariate drift calculator are stored in the uv_results
variable.
This exercise is part of the course
Monitoring Machine Learning in Python
Exercise instructions
- Set period argument to
analysis
fordrift_results
. - Pass hotel and country to
column_names
fordrift_results
. - Set
kind
argument in.plot()
method to"drift"
. - Do the same for
distribution_results
, except for setting thekind
argument in.plot()
method to"distribution"
.
Hands-on interactive exercise
Have a go at this exercise by completing this sample code.
# Filter and create drift plots
drift_results = uv_results.filter(
period=____,
column_names=[____, ____]
).plot(kind=____)
# Filter and create distribution plots
distribution_results = uv_results.filter(
period=____,
column_names=[____, ____]
).plot(kind=____)
# Show the plots
drift_results.show()
distribution_results.show()