Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 20 (Information that should be communicated during a presentation,…
Chapter 20
Information that should be communicated during a presentation
Business problem
This should be used to guide all work
Model quality metrics (confusion matrix)
Introduce in presentation without orange add ons, focus on measure rather than name
Try to use simple graphs
Areas where a model struggles
Internal data- talk about where internal data could be improved to make the model better
External data- maybe purchase certain data within the budget that could also help
most predictive features for model building
Not all predictive features are of equal importance
Four features to consider before bringing to management
Features that need to be changed which would force you to need to re-run the models
Features that need further resarch, could use the help of a subject matter expert
Immutable feauters- good for modeling but not helpful to management to implement corrective action.
This is a feature that management cannot change
Mutable features- features that management could possbly change. Could be manipulated to improve data set
However, anytime management changes stuff the model loses efficiacy. With more improvements though this is just sort of a reality and could ultimately improve the model
feature types especially interesting to management
recommend business actions
Make sure there aren't any mistakes in the model creation process before presenting
Unlocking holdout
Hold out sample gives an opportutnity to check if problems hve occured