CharlotteJackson/DC_Crash_Bot

Given an address, determine most unsafe time

Closed this issue · 3 comments

banjtheman commented

What is the Task

Given an address return the time period which the most incidents

Why do we want to do this

Step 3 in the TSA is "Days and time when safety concerns are the worst:"
(Such as weekday AM peak, weekday PM peak, overnight, weekends, etc.)

How can I get started?

Would start by evaluating the crash datasets, and see if you can pull out the time and day of most incidents given an address
https://docs.google.com/spreadsheets/d/18OQh78KhlL65JHM2DyQKVEdYUY8zXjHzTfYbdsd-Kes/

Here is a subset in the repo
https://github.com/CharlotteJackson/DC_Crash_Bot/blob/master/data/analysis_data_dc_crashes_w_details.csv

Definition of Done

A function that can take an address as input and return a what day, and time has the most incidents

banjtheman commented

Was able to do the following

  • Create a reverse geocode function that gives a lat-long given an address
  • with the lat long, find all crashes that happened within .2 miles
  • Add the crash to a bucket of what time of day it happened, and on weekend or weekday
  • Print out results

Example below

image

banjtheman commented

Questions to bring up in next meeting

  • What threshold should we use to determine if a crash is close enough (currently using .2)
  • Is one year good enough to collect data on?
  • Are we okay with the text generated?