Problem
Problem
Common-area charges are not recovered evenly across locations. A score fit on the same year it is judged flatters the result.
Case study
CAM recovery likelihood
Design
Design
Train only on earlier years. Hold out one test year. Judge how cleanly recoveries separate from the rest in that unseen year.
Implementation
Implementation
Run one walk-forward pass per year from 2009 through 2024. Fit on the prior years only. Record a separation score for that test year.
Benefit
Benefits
Each bar is a year the model had not seen. A taller bar is a year when recoveries and non-recoveries were easier to tell apart.