Upload
geco-in-the-rockies
View
45
Download
0
Embed Size (px)
Citation preview
Geocoding Employment Data Implementing Different Methods and
Why Precision and Accuracy Matter
56 Member Jurisdictions
10 Counties
46 Municipalities
5288 Square Miles
Urban Sim Model
2010 Base year
949,484 Parcels
881,751 Buildings
1,692,409 Jobs
Purchased Regional Employment Data
140,809 records
Lat/long provided, but method unknown/proprietary
Is the data spatially correct?
Compare same data, different geocode method
QUESTION:
WHAT IS THE BEST WAY
TO QUICKLY GEOCODE
ALL THIS DATA?
Repeatable
Accurate
Precise
Applicable to other data
HYPOTHESIS
•Denver Address Dataset - Denver City/County
•Google Geocode - DRCOG Region
•Lat/Long from data firm- DRCOG Region
•Composite Parcel addresses - DRCOG Region
Methodology: Denver Address Dataset
(DAD)
Download FREE Denver Address Dataset from Denver’s
Open Data Catalog
Address Locator with Denver Address Dataset
Geocode all Employment Data with DAD Locator
Methodology: Google
.xls GeokettlePostgreSQL (pgAdmin)
Python script runs
Google API
◦ lat/long
Spatial function to
create location
Methodology: Parcel Addresses
Create address locator for every county
Composite address locator
Geocode all employment data with County Composite
Arapahoe
Adams
Boulder
Broomfield
Clear
Creek
Denver
Douglas
Gilpin
Jefferson
Weld
Results: The Good… DAD
Results: The Good & The Bad… Google
Results: The Good & The Bad… Proprietary
Lat/Long
Results: …The Ugly Composite Parcel Locator
Results: Raw Numbers for Region
Out of 140,809 establishments and
1,604,052 jobs
Google: 113,613 establishments
◦ 1,316,241 jobs
Parcel: 61,555 establishments
◦ 710,822 jobs
Why does it matter?
Areas of importance
◦ Urban Centers
Establishments Jobs
Proprietary
Lat/Long
14,693 211,449
DAD 7,197 96,116
Google 12,127 191,412
Parcel 5,552 75,224
Denver City/County Urban Centers
Why does it matter?
Areas of importance
◦ ½ Mile Transit Buffers
Denver City/County Fastrack Stations
Establishments Jobs
Proprietary
Lat/Long
12,294 188,889
DAD 6,689 89,108
Google 9,709 146,873
Parcel 5,316 89,108
Conclusions QAQC is difficult!
Proprietary Lat/Long is best for region
Repeatable?
Can it be used in conjunction with other data?
DAD is best for Denver City/County
Google has limitations in the region
Expensive $10K/year
100,000 records a day
Parcel address locator is not a viable geocoding method
Quickest, data easiest to obtain
Too many missing addresses
Too many incomplete addresses
NEXT STEPS More municipalities to track address data
Apply this to other employment data
QCEW
Change address locator and geocoding default
settings