Updated address normalization

This commit is contained in:
2025-12-15 15:30:45 -05:00
parent cfd8cfb57f
commit d234e366bd
2 changed files with 225 additions and 7 deletions
+208
View File
@@ -0,0 +1,208 @@
"""
Address Normalization Fix for Distance Calculation Issues
==========================================================
This module fixes the issue where nearly identical addresses are geocoded to
different coordinates, causing incorrect distance calculations.
Issue:
- "7100 Gordon Rd" vs "7100 Gordons Rd, USA" → 1.5 miles apart (WRONG!)
- "3402 South Glebe Road" vs "3402, South Glebe Road, Aurora Hills" → 0.7 miles (WRONG!)
Root Cause:
- Google Maps/OSM geocodes slightly different address strings to different coordinates
- Minor variations (plurals, commas, neighborhoods, "USA") create false distance
Solution:
- Normalize addresses before geocoding
- Use fuzzy matching to detect identical locations
- Prevent re-geocoding of essentially the same address
"""
import re
from difflib import SequenceMatcher
def normalize_address(address):
"""
Normalize address string for better matching and geocoding accuracy
This helps prevent geocoding nearly identical addresses to different coordinates
Args:
address: Raw address string
Returns:
Normalized address string
Examples:
"7100 Gordon Rd, Falls Church, VA 22043"
"7100 Gordons Rd, Falls Church, VA 22043, USA"
Both normalize to: "7100 gordon rd, falls church, va 22043"
"3402 South Glebe Road Arlington VA 22202"
"3402, South Glebe Road, Aurora Hills, Arlington VA 22202"
Both normalize to: "3402 s glebe rd, arlington, va 22202"
"""
if not address or not isinstance(address, str):
return address
# Convert to lowercase for consistent comparison
normalized = address.lower().strip()
# Remove extra whitespace and normalize separators
normalized = re.sub(r'\s+', ' ', normalized) # Multiple spaces to single space
normalized = re.sub(r'\s*,\s*', ', ', normalized) # Normalize comma spacing
# Standardize common street abbreviations to short forms
street_abbrev = {
r'\broad\b': 'rd',
r'\broads\b': 'rd', # Handle plural form (Gordon Rd vs Gordons Rd)
r'\bstreet\b': 'st',
r'\bavenue\b': 'ave',
r'\bdrive\b': 'dr',
r'\blane\b': 'ln',
r'\bcourt\b': 'ct',
r'\bboulevard\b': 'blvd',
r'\bparkway\b': 'pkwy',
r'\bcircle\b': 'cir',
r'\bplace\b': 'pl',
r'\bterrace\b': 'ter',
r'\bhighway\b': 'hwy'
}
for full_form, abbrev in street_abbrev.items():
normalized = re.sub(full_form, abbrev, normalized)
# Standardize directionals to single letter
directionals = {
r'\bnorth\b': 'n',
r'\bsouth\b': 's',
r'\beast\b': 'e',
r'\bwest\b': 'w',
r'\bnortheast\b': 'ne',
r'\bnorthwest\b': 'nw',
r'\bsoutheast\b': 'se',
r'\bsouthwest\b': 'sw'
}
for full_form, abbrev in directionals.items():
normalized = re.sub(full_form, abbrev, normalized)
# Remove neighborhood/district names that aren't essential for location
# Examples: "Aurora Hills", "Downtown", etc.
parts = [p.strip() for p in normalized.split(',')]
# Keep: street address, city, state, zip
# Remove: neighborhood names, building names, country suffixes
filtered_parts = []
# Known neighborhood keywords to remove (these don't affect geocoding)
neighborhood_keywords = ['hills', 'heights', 'park', 'village', 'estates',
'manor', 'gardens', 'terrace', 'commons', 'plaza']
for i, part in enumerate(parts):
part_clean = part.strip()
# Always keep first part (street address)
if i == 0:
filtered_parts.append(part_clean)
continue
# Skip empty parts
if not part_clean:
continue
# Skip country suffixes
if part_clean in ['usa', 'us', 'united states']:
continue
# Skip if it's a neighborhood name (contains neighborhood keywords but no numbers)
is_neighborhood = False
for keyword in neighborhood_keywords:
if keyword in part_clean and not re.search(r'\d', part_clean):
is_neighborhood = True
print(f" Removing neighborhood: '{part_clean}'")
break
if is_neighborhood:
continue
# Keep if it looks like state (2 letter abbrev)
if re.match(r'^[a-z]{2}$', part_clean):
filtered_parts.append(part_clean)
continue
# Keep if it looks like zip code
if re.match(r'^\d{5}(-\d{4})?$', part_clean):
filtered_parts.append(part_clean)
continue
# Keep if it's likely a city name (reasonable length, no special patterns)
if 3 <= len(part_clean) <= 30:
filtered_parts.append(part_clean)
# Reconstruct address
normalized = ', '.join(filtered_parts)
# Remove common country suffixes that don't affect location
normalized = re.sub(r',?\s*(usa|united states|us)$', '', normalized)
# Final cleanup: remove trailing commas and spaces
normalized = normalized.strip(', ')
print(f"🔧 Address normalization:")
print(f" Original: {address}")
print(f" Normalized: {normalized}")
return normalized
def addresses_are_similar(addr1, addr2, threshold=0.90):
"""
Check if two addresses are similar enough to be considered the same location
Uses fuzzy string matching to handle minor variations
Args:
addr1: First address string
addr2: Second address string
threshold: Similarity threshold (0-1), default 0.90 (90% similar)
Returns:
Boolean indicating if addresses are similar
Examples:
addresses_are_similar(
"7100 Gordon Rd, Falls Church, VA 22043",
"7100 Gordons Rd, Falls Church, VA 22043, USA"
) → True (same location, minor spelling difference)
addresses_are_similar(
"3402 South Glebe Road Arlington VA 22202",
"3402, South Glebe Road, Aurora Hills, Arlington VA 22202"
) → True (same location, extra neighborhood name)
"""
if not addr1 or not addr2:
return False
# Normalize both addresses
norm1 = normalize_address(addr1)
norm2 = normalize_address(addr2)
# Exact match after normalization
if norm1 == norm2:
print(f"✅ Addresses match exactly after normalization")
return True
# Calculate similarity using difflib SequenceMatcher
similarity = SequenceMatcher(None, norm1, norm2).ratio()
is_similar = similarity >= threshold
print(f"📊 Address similarity check:")
print(f" Address 1 (normalized): {norm1}")
print(f" Address 2 (normalized): {norm2}")
print(f" Similarity score: {similarity:.2%}")
print(f" Threshold: {threshold:.2%}")
print(f" Result: {'✅ SIMILAR (same location)' if is_similar else '❌ DIFFERENT (different locations)'}")
return is_similar