Please enable JavaScript.
Coggle requires JavaScript to display documents.
Communicate Model Insights (make sure your model is relevant to your…
Communicate Model Insights
make sure your model is relevant to your audience
5 types of information to be communicated during a presentation
never enough practice to make you a perfect presenter or information
Business problem
Model quality metrics
Most predictive features
feature types interesting to management
recommended business actions
Unlocking Holdout
check to make sure no mistakes were made by unlocking the holdout
once it is unlocked check to see if the ranking of your algorithms changed
small holdout samples will likely lead to more issues
Business problem first
keep the same business problem in mind that began your research
Pre-processing and Model Quality Metrics
one quality metric is the confusion matrix
note the models positive predictive value
Areas where the model struggles
argue a pilot for the project
internal data --> collect more data within the organization
External data --> data from outside the organization
Most Predictive Features
develop a story for what features you took out and which ones are left
not all features will be interesting to your audience
Not all features are created equally
4 features to consider before your presentation
Features that need to be changed and therefore require a re-run
Features requiring further examination
Immutable features
Mutable features
Recommend business actions
final recommendations