我有一个地址匹配算法，该算法将一个客户ID的地址的密码与另一个客户ID的密码进行比较，并创建具有类似地址类型的客户ID的列表。接下来，我为每个客户ID列表分配一个特定的组号。但是问题是，大多数客户ID列表都按适当的组号分组，但是即使某些ID具有相同的地址，它们也按不同的数字分组。
我制作了地址匹配脚本，找到了具有相似地址的客户ID，并将其放在列表中，并为每个列表分配了一个组号。但是，即使在地址相似且位于同一列表中的组号不同之后，组号对于某些ID和某些ID也是正确的
import pandas as pd
from fuzzywuzzy import fuzz
from fuzzywuzzy import process
def how_similar(address,address_list):
address = address.strip()
address1_split = address.split(" ")
pinCode_address1 = address1_split[-1]
area_address1 = " ".join(address1_split[:-1])
matched_address = []
for add in address_list:
score = 0
add = add.strip()
address2_split = add.split(" ")
pinCode_address2 = address2_split[-1]
area_address2 = " ".join(address2_split[:-1])
area_score = fuzz.token_set_ratio(area_address1,area_address2)
score+=area_score
if score>= 90 :
matched_address.append(add)
return matched_address
def pat_match(df):
# Getting the column values of id and address in seprate list
id = df['COD_CUST_ID'].values.tolist()
address = df['ADDRESS'].values.tolist()
# Creating a new column with name 'Ids'
df['Ids'] = ""
length01=len(id)
for y in range(0,length01):
#print(y)
# The mathched address Id will will be appended in a list for every address
matched_ids = []
# Calculating list of address with match percentage more than 80%
#score=process.extractBests(address[y],address,score_cutoff=80)
score= how_similar(address[y],address)
# Iterating over every address returned by score one by one
for matched_address in score:
# Getting Customer_ID of every Address
get = df['COD_CUST_ID'][df['ADDRESS']==matched_address].tolist()
# Appending the Id into a list
if get not in matched_ids:
matched_ids.append(get)
# Finally Appending the list of matched ID to the column
df['Ids'][df['COD_CUST_ID']==id[y]] = str(matched_ids)
# Creating a new column with name 'Group' and 'Matched'
df['group'] = ""
df['Matched'] = ""
grouped_customers = []
Group_count = 1
for i in range(length01):
try:
id_list = df['Ids'][i]
new_list = id_list[1:-1]
new_list=new_list.replace('[','')
new_list=new_list.replace(']','')
new_list = new_list.split(',')
new_list = [int(x) for x in new_list]
#print(new_list)
# print(i)
for id in new_list:
if id not in grouped_customers:
df['group'][df['COD_CUST_ID'] == id] = Group_count
df['Matched'][df['COD_CUST_ID'] == id] = id
grouped_customers.append(id)
if id == new_list[-1]:
Group_count+=1
except Exception as e:
print(str(e))
df.to_csv('match_score_8.csv',sep=',',index=None)
if __name__ == '__main__':
data = pd.read_csv('address.csv')
pat_match(data)
假设我的输出为
COD_CUST_ID ADDRESS Ids group Matched
23656386 VPO- DHANGER KHER ABOHAR 152116 [[23656386], [23656388, 23656387]] 8 23656386
23656388 VPO- DHANGER KHERA ABOHAR 152116 [[23656386], [23656388, 23656387]] 8 23656388
23656387 VPO- DHANGER KHERA ABOHAR 152116 [[23656386], [23656388, 23656387]] 8 23656387
这是大多数ID所需的实际输出。但是有一些ID的输出就像
COD_CUST_ID ADDRESS Ids group Matched
23656887 TODAWATA KI DHANI FATEHPURA BANSA 303806 [[23656887], [23666683], [23666717, 23666721, 23666773]] 416 23656933
23666683 - - FATEHPURA BANSA 303806 [[23656980], [23656887], [23666683], [23666717, 23666721, 23666773]] 196 23666664
23666717 BANSA 303806 [similar list of matching ids] 151 23666717
23666721 BANSA 303806 [similar list of matching ids] 151 23666721
23666773 BANSA 303806 [similar list of matching ids] 151 23666773
匹配的ID应该在同一组中