问题描述
我有一个Pandas数据框，该数据框收集进行交易的供应商的名称.由于此数据是自动从银行对帐单中收集的，因此许多供应商都是相似的……但并不完全相同.总之，我想用一个名称替换供应商名称的不同排列.
I have a Pandas dataframe that collects the names of vendors at which a transaction was made. As this data is automatically collected from bank statements, lots of the vendors are similar... but not quite the same. In summary, I want to replace the different permutations of the vendors' names with a single name.
我认为我可以找到一种方法来做到这一点(见下文)，但是我是一个初学者，在我看来这是一个复杂的问题.我真的很想知道更多有经验的编码人员将如何使用它.
I think I can work out a way to do it (see below), but I'm a beginner and this seems to me like it's a complex problem. I'd be really interested to see how more experienced coders would approach it.
我有一个这样的数据框(在现实生活中，它大约有20列，最多约50行):
I have a dataframe like this (in real life, it's about 20 columns and a maximum of around 50 rows):
Groceries Car Luxuries
0 Sainsburys Texaco wst453 Amazon
1 Sainsburys bur Texaco east Firebox Ltd
2 Sainsbury's east Shell wstl Sony
3 Tesco Shell p/stn Sony ent nrk
4 Tescos ref 657 Texac Amazon EU
5 Tesco 45783 Moto Amazon marketplace
我想找到类似的条目，并用这些条目的第一个实例替换它们，所以我最终得到了这一点:
I'd like to find the similar entries and replace them with the first instance of those entries, so I'd end up with this:
Groceries Car Luxuries
0 Sainsburys Texaco wst453 Amazon
1 Sainsburys Texaco wst453 Firebox Ltd
2 Sainsburys Shell wstl Sony
3 Tesco Shell wstl Sony
4 Tesco Texaco wst453 Amazon
5 Tesco Moto Amazon
我的解决方案可能远非最佳.我当时想按字母顺序排序，然后按位排序，并使用difflib中的SequenceMatcher之类的东西来比较每对供应商.如果相似度高于某个百分比(我希望一直使用此值直到满意为止)，然后将假定这两个供应商是相同的.我担心我可能正在使用大锤敲碎螺母，或者可能会花费很长时间(我并不迷恋性能，但同样地，我也不想等待数小时).
My solution might be far from optimum. I was thinking of sorting alphabetically, then going through bitwise and using something like SequenceMatcher from difflib to compare each pair of vendors. If the similarity is above a certain percentage (I'm expecting to play with this value until I'm happy) then the two vendors will be assumed to be the same. I'm concerned that I might be using a sledgehammer to crack a nut, or it might take a long time (I'm not obsessed with performance, but equally I don't want to wait hours for the result).
真的很想听听人们对这个问题的想法！
Really interested to hear people's thoughts on this problem!
推荐答案
在开始时，问题似乎并不复杂，但是确实如此.我不喜欢我的代码，必须有更好的方法.但是我的代码正在工作.
At the start, the problem doesn't seem complicated, but it is. I didn't like my code, there must be a better way. However my code is working.
我使用名为 fuzzywuzzy 的字符串相似性软件包来确定必须替换的字符串.该软件包使用Levenshtein相似性，我使用％90作为阈值.另外，任何字符串的第一个单词都用作比较字符串.这是我的代码:
I used string similarity package named fuzzywuzzy to decide which string must be replaced. This package uses Levenshtein Similarity, and I used %90 as threshold value. Also, first word of any string is used as comparison string. Here is my code:
import pandas
from fuzzywuzzy import fuzz
# Replaces %90 and more similar strings
def func(input_list):
for count, item in enumerate(input_list):
rest_of_input_list = input_list[:count] + input_list[count + 1:]
new_list = []
for other_item in rest_of_input_list:
similarity = fuzz.ratio(item, other_item)
if similarity >= 90:
new_list.append(item)
else:
new_list.append(other_item)
input_list = new_list[:count] + [item] + new_list[count :]
return input_list
df = pandas.read_csv('input.txt') # Read data from csv
result = []
for column in list(df):
column_values = list(df[column])
first_words = [x[:x.index(" ")] if " " in x else x for x in column_values]
result.append(func(first_words))
new_df = pandas.DataFrame(result).transpose()
new_df.columns = list(df)
print(new_df)
输出:
Groceries Car Luxuries
0 Sainsbury's Texac Amazon
1 Sainsbury's Texac Firebox
2 Sainsbury's Shell Sony
3 Tesco Shell Sony
4 Tesco Texac Amazon
5 Tesco Moto Amazon
我想，func()函数可以更好地编码，但这是我首先想到的.
I guess, func() function can be coded better but this is what comes first to my mind.
这篇关于用相同的字符串替换列中的相似字符串的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持！