09 · Personal & Academic Project
Behavioral Risk & Bed Demand
Protecting population patterns while forecasting regional capacity
Business value
Behavioral indicators were translated into interpretable regional demand and capacity-strain signals.
01 · Business problem
212,000+ records included more than 10,000 missing values. Naive imputation risked distorting population-level patterns.
02 · My approach
I designed cleaning around preserving distributions, then compared linear and nonlinear models for demand forecasting.
03 · What I built
Structured imputation, Gradient Boosting, Ridge Regression, and Tableau dashboards.
What changed my thinking
The corrections, tradeoffsand decisions behind it.
- 01
Missing data was a modeling decision
The imputation method could materially change the population story.
- 02
Two model families improved judgment
Ridge supported interpretation while Gradient Boosting captured nonlinear patterns.
- 03
The dashboard completed the workflow
Forecasts became useful only after they were translated into planning views.