快速模糊匹配合并

Question

对此非常陌生，如果有以下建议，我将不胜感激：

我有一个数据集'Projects‘，显示具有项目ID的机构列表：

project_id  institution_name
0           somali national university
1           aarhus university
2           bath spa
3           aa school of architecture
4           actionaid uk

我想将其与以下数据集“大学”及其国家代码进行模糊匹配合并：

institution_name                      country_code
a tan kapuja buddhista foiskola             HU
aa school of architecture                   UK
bath spa university                         UK
aalto-yliopisto                             FI
aarhus universitet                          DK

然后找回这个：

project_id  institution_name           Match    organisation               country_code
0           somali national university []       NaN                        NaN
1           aarhus university          [(91)]   aarhus universitet         DK
2           bath spa                   [(90)]   bath spa university        UK
3           aa school of architecture  [(100)]  aa school of architecture  UK
4           actionaid uk               []       NaN                        NaN

使用rapidfuzz：

import pandas as pd

import numpy as np

from rapidfuzz import process, utils as fuzz_utils

def fuzzy_merge(baseFrame, compareFrame, baseKey, compareKey, threshold=90, limit=1, how='left'):
    #   baseFrame: the left table to join
    #   compareFrame: the right table to join
    #   baseKey: key column of the left table
    #   compareKey: key column of the right table
    #   threshold: how close the matches should be to return a match, based on Levenshtein distance
    #   limit: the amount of matches that will get returned, these are sorted high to low
    #   return: dataframe with boths keys and matches
    s_mapping = {x: fuzz_utils.default_process(x) for x in compareFrame[compareKey]}

    m1 = baseFrame[baseKey].apply(lambda x: process.extract(
      fuzz_utils.default_process(x), s_mapping, limit=limit, score_cutoff=threshold, processor=None
    ))
    baseFrame['Match'] = m1

    m2 = baseFrame['Match'].apply(lambda x: ', '.join(i[2] for i in x))
    baseFrame['organisation'] = m2

    return baseFrame.merge(compareFrame, on=baseKey, how=how)

Merged = fuzzy_merge(Projects, Universities, 'institution_name', 'institution_name')

Merged

我得到了这个(在match列中有一些额外的文本，但现在不会深入讨论)。这几乎就是我想要的，但是国家代码只有在100%匹配时才会匹配：

project_id  institution_name           Match    organisation               country_code
0           somali national university []       NaN                        NaN
1           aarhus university          [(91)]   aarhus universitet         NaN
2           bath spa                   [(90)]   bath spa university        NaN
3           aa school of architecture  [(100)]  aa school of architecture  UK
4           actionaid uk               []       NaN                        NaN

我认为这是我如何将我的basekey与compareframe进行比较以创建我的合并数据集的问题。不过，我不知道如何在“组织”上返回它--尝试插入它会导致不同的错误。

Answer 1

不要紧，我想明白了--我没有考虑到空单元格！用NaN替换它们效果很好。

def fuzzy_merge(baseFrame, compareFrame, baseKey, compareKey, threshold=90, limit=1, how='left'):
    s_mapping = {x: fuzz_utils.default_process(x) for x in compareFrame[compareKey]}

    m1 = baseFrame[baseKey].apply(lambda x: process.extract(
      fuzz_utils.default_process(x), s_mapping, limit=limit, score_cutoff=threshold, processor=None
    ))
    baseFrame['Match'] = m1

    m2 = baseFrame['Match'].apply(lambda x: ', '.join(i[2] for i in x))
    baseFrame['organisations'] = m2.replace("",np.nan)

    return baseFrame.merge(compareFrame, left_on='organisations', right_on=compareKey, how=how)