Please enable JavaScript.
Coggle requires JavaScript to display documents.
Predictive modeling (Competitive advantage (Streamlined and less $ UW…
Predictive modeling
Competitive advantage
In-force mgmt
Surrender detection
emerging mortality
Predict interest in buying life ins.
More efficient marketing
identify most qualified customers
cross-selling
ID highest yielding population
health risk
likeliness to buy or retain
Streamlined and less $ UW
shorter process
handle larger volume
reduce medical test costs
triage the application
=> immediate pass
or further review
Improve efficiency of agent training
score prospective agents
Prerequisite
Suitable data
large # of observations
Clearly defined target variable
Application to convert output to actions
Steps to build model
prepare data
Exploratory data analysis => distributional attributes
Var transformation
replace missing values
removal (not ideal)
replace with best estimate
reduce outliers
take log
cap extreme value
group into buckets => higher credibility
convert into numeric values to capture trends
Variable Generation => create var from raw data
Partition data into 3 sets
Validation set => initial calibration to trained model
Test set => final testing of model
Training set => initial modeling
build model
stepwise regression
principal components analysis
Train the model
Validate model
Test model
Collect data
final adjustment
Target variable
Mortality (concerns)
Low claim frequency
Low data quality
data volume is low
UW decisions
Advantage
diversification
quick feedback
expert judgment
Disadvantage
Inconsistency in human decision
company-wide bias in UW
anti-selection