Example of using Mapzen Search API with prioritization for geocoding

README.md

Geocoding with Mapzen Search and using geographic prioritization

The Mapzen Search API is built from open-source data and technology and allows up to 30,000 geocoding requests per day. This gist shows how to use the Prioritize within a circular region feature to get more accurate results.

Here's the results of geocoding the Starbucks locations in New York City, using the priorization feature:

The data files in this gist:

• starbucks_geocoding.mapzen.priority.py - a Python script that reads starbucks.csv and creates starbucks.geocoded.csv
• starbucks.csv - a list of Starbucks locations as derived from the city's health inspection database
• starbucks.geocoded.csv - the results of running the Python script with Mapzen geocoding.

The API call, with parameters to restrict the search to a focus point and boundary, looks like this:

  /v1/search?api_key=search-XXXXXXX
     &text=YMCA
     &focus.point.lat=-33.856680
     &focus.point.lon=151.215281
     &boundary.circle.lat=-33.856680
     &boundary.circle.lon=151.215281
     &boundary.circle.radius=50

Without using these parameters to limit the scope of the search, Mapzen search will sometimes return erroneous results.

For example, the top result of 1 PENN PLAZA MANHATTAN, NY, 10119 will be geocoded to Kansas:

However, specifying a boundary circle around New York, and the erroneous points will (mostly) be fixed:

There's still a few errors...for example, 0 JFK DELTA AIRLINES, QUEENS, NY 11422 is geocoded to be near La Guardia Airport.

In my Python script, I create a lookup table for each borough (e.g. BRONX, NY), and for each result, I set the focus point to the borough point. There's probably better results to be found by creating a lookup table for each zipcode.

See the CartoDB version of these maps:

• With no prioritization (i.e. take the first result)
• With boundary prioritization

starbucks.csv

starbucks.geocoded.csv

starbucks_geocoding.mapzen.priority.py

from csv import DictReader, DictWriter
from os import environ
from time import sleep
import requests
FNAME = 'starbucks.csv'
ONAME = 'starbucks.geocoded.csv'
API_KEY = environ['MAPZEN_SEARCH_KEY']
BASE_URL = 'https://search.mapzen.com/v1/search'
# What I got when I googled for lat/lng for "New York City"
NEW_YORK_LAT = 40.7306
NEW_YORK_LON = -73.935

# Using the priority boundary feature in Mapzen Search
# https://mapzen.com/documentation/search/search/#prioritize-within-a-circular-region
def foo_ny_geocode(address_text, 
                    focus_point = {'lat': NEW_YORK_LAT ,'lon': NEW_YORK_LON}):
  resp = requests.get(BASE_URL, 
                    params = {'api_key': API_KEY, 'size': 1, 
                      'text': address_text, 
                      'focus.point.lat': focus_point['lat'],
                      'focus.point.lon': focus_point['lon'],
                      'boundary.circle.lat': focus_point['lat'],
                      'boundary.circle.lon': focus_point['lon'],
                      'boundary.circle.radius': 80})
  data = resp.json()
  bbox = data['bbox']  
  pt = {'distance_from_focus': data['features'][0]['properties']['distance']}
  pt['lat'] = (bbox[1] + bbox[3]) / 2
  pt['lon'] = (bbox[0] + bbox[2]) / 2
  return pt


# make a lookup table for all boros
BORO_POINTS = {boro: foo_ny_geocode(boro + ", NY") for boro in ['QUEENS', 'MANHATTAN', 'BRONX', 'STATEN ISLAND', 'BROOKLYN']}

r = open(FNAME)
rcsv = DictReader(r)
w = open(ONAME, 'w')
wcsv = DictWriter(w, fieldnames = rcsv.fieldnames + ['latitude', 'longitude'])
wcsv.writeheader()
for row in rcsv:
  addr = "%s %s, %s, NY %s" % (row['BUILDING'], row['STREET'], row['BORO'], row['ZIPCODE'])
  # 140 E 45TH ST, MANHATTAN, NY 10017
  try:
    pt = foo_ny_geocode(addr, focus_point = BORO_POINTS[row['BORO']])
  except Exception as err:
    print(addr)
    print("\t", err)
  else:    
    print(addr, pt['distance_from_focus'], pt['lat'], pt['lon'])
    row['latitude'] = pt['lat']
    row['longitude'] = pt['lon']
    wcsv.writerow(row)
    sleep(0.2)

w.close()
r.close()