Additional households fixes based on testing #58

dehall · 2023-04-20T17:31:14Z

Ensure the "written_to_file" flag is read properly
Tweak match thresholds to reduce tumbleweeds, in particular in apartments
Fix for referencing a None object
Adds a new --exact_addresses option
- This uses a faster but lower quality method for address comparison. By default the inference process will split up addresses into street, number, suffix, etc, and use family name and phone number, but this increases memory & runtime. Enabling this feature causes the process to use the entire address as a single string for comparisons, and only the address is considered. If addresses have not been standardized/validated, this setting will increase false negatives (records not being included in households where they should be). This is primarily intended to serve as a baseline for testing the software, but might be used in practice as a last resort.
Prints some basic household summary stats at the end
Also adds an extra print out at the end of linkid_to_patid of the name of the filename that was written

additional fixes based on testing

f0d87d5

dehall changed the title ~~Additional fixes based on testing~~ Additional households fixes based on testing Apr 20, 2023

dehall added 6 commits April 21, 2023 10:59

add fast_addresses option

60e9548

write some basic household stats at the end

c8f87dd

return threshold back down to 0.85

5715a5f

style fixes

c5f05e8

tweak thresholds further and improve exact_addresses feature

a881061

print out the filename that was written

4b9256a

dehall merged commit 6ec0195 into master Apr 26, 2023

dehall deleted the further_household_fixes branch April 26, 2023 13:53

Provide feedback