Please enable JavaScript.
Coggle requires JavaScript to display documents.
Chapter 20: Communicate Model Insights (Most Predictive Features…
Chapter 20: Communicate Model Insights
1.Unlocking Holdout
Make sure No mistakes in model
Go to leaderboard and click
Unlock
Observe
Cross Validation
and
Holdout
Best outcome are ones in which the order of models did not change between the two sorts
Second Best: At least the top models stay at the top of the list
2.Business Problem First
Updated Problem Statement
We will create a
machine-learning model capable of
detecting which patients are likely to be readmitted within 30 days of discharge and develop educational and support programs targeting these patients. 60% of our patients are discharged to home without home health services;
this group mirrors the overall patient population in terms of average readmission.
We are especially interested in detecting cases where these patients are highly likely to be readmitted.
3.Pre-processing and Model Quality Metrics
Presentables:
Confusion Matrix
Annotate
Explain model’s performance AKA:
Positive Predictive Value
True Positive Rate
Overview explaining the process for procuring data, cleaning that data, and carefully addressing issues
Confusion matrix
with a different probability distribution threshold
Got to
Predict
Add the
discharge disposition id
and
readmitted
through
the Optional Features
Select
Holdout
, then "Compute Predictions"
Areas where the model struggles
Internal Data
More data on past visits should be collected for patients that are
frequent visitors
Mine patients record for additional data
External Data
Purchased Data
Example: Grocery Store Data, financial data
Most Predictive Features
directional results tend to make intuitive sense
Knowing every value isn't necessary
Develop story around findings based on the extensive examination of features
Not All Features are Created Equally
4 Types of features to consider before going into a presentation
Features requiring further examination
Features about which a manager is likely to ask follow-up questions
3.
Immutable features
Features that are good for modeling but are of no value to management
1.
Features that need to be changed and therefore require a re-run of the models
General rule: When features need to be changed, its necessary to re-run the model after making changes
Mutable features
Features that management could potentially change
If changes do occur, model loses efficiency
Recommended Business Actions
Explicit considerations for next steps