Calgary’s Inundation Model and Its Application to Project Denver’s Flood Risks

Click for Video Presentation:

Overview

This exercise applied geo-spatial machine learning to learn the inundation model in Calgary and predicted flooding hazard in another geography that is the city of Denver.

Motivation

To adopt a scientific approach for studying and predicting natural disasters; to assist policy makers in the environment and infrastructure section to formulate adaptive strategies; and to protect assets and finance of local residents.

Methodology & Algorithm

The data analysis exercise will take the following steps:

• Determining a city for comparison: Denver
  
• Collecting data for Calgary and Denver
  
• Feature engineering
  
• Create fishnet
  
• Forward and Backward Selection for non-multicollinear criteria
  
• Logistic regression for significant factors
  
• Training vs Testing
  
• Model validation & cross validation
  
• Prediction on Denver
  

Variables

Confusion Matrix and ROC Curve

Prediction with Calgary Test Geography and Denvor