Optimize ambulance placement







Roy Hyunjin Han
March 27, 2014


Mountain View Spatiotemporal Analysis Working Group Meetup

Overview



Goal
Methods
Techniques
Discussion






Press the right arrow key to skip to the next topic
Press the down arrow key to explore a topic

Goal




Reduce emergency response time



Place ambulances close to where emergencies might happen





From Benjamin Dean of the NYC's Mayor's Office of Data Analytics

Reduce emergency response time




Estimate where emergencies might happen

using past 911 calls


Optimize ambulance placement

Optimize ambulance placement




Generate ranked list of recommended ambulance locations

to maximize number of past calls within 800m of an ambulance

and minimize summed distance of past calls to nearest ambulance

Method


Random
Single
Split
Rank

Place five ambulances at random

 

Place one ambulance


Divide and conquer


 

Survival of the fittest


 

Method comparison



Technique


K-D tree

Brute force gradient descent

K-D tree


kdtree = KDTree(points)

kdtree.query(
    x,
    maximum_count,
    maximum_distance)

Brute force gradient descent


scipy.optimize.fmin_bfgs(f, x0)



 

Discussion


Form groups of three

Think of your approach

Present approach