Optimized for currently employed customers using a multi-model ensemble (XGBoost + Random
Forest + LightGBM) to capture complex patterns.
Retired Model
Specifically tuned for the retired demographic using an optimized XGBoost model, focusing
on stability and long-term loyalty factors. More features so less complex model with
same prediction power
Churn vs. Retention: The model predicts the probability of a policyholder cancelling
or lapsing their policy at the next renewal date.
0 (Retain): Policy renewed.
1 (Churn): Policy cancelled or lapsed.
Observation Window
The model utilizes customer behavioral data, claim history,
and policy modifications from the previous 3 years leading up to the renewal
decision point.
Prediction Window
The prediction is valid for the upcoming renewal cycle (12 months).