如果完整字符串包含在另一个pandas datafram中

Question

我想用数据帧对部件进行分类。你知道吗

简化问题以尝试显示问题：

data = {'col1': ['engine','blue engine cover','spark plug',
        'rear panel','black rear panel', 'blue engine']}
desc_df = pd.DataFrame(data=data)

catg = {'bodywork': ['engine cover','side panel','rear panel'],'underhood':['engine','spark plug','oil filter'],
   'Glass':['Windscreen','window','demister']}

catg_df = pd.DataFrame(data=catg)

catg_df


   Glass         bodywork       underhood
0 Windscreen     engine cover   engine 
1 window         side panel     spark plug 
2 demister       rear panel     oil filter 

desc_df

     col1
0   engine 
1 blue engine cover 
2 spark plug 
3 rear panel 
4 black rear panel 
5 blue engine 

最后我想说：

  col1                Category
0 engine              underhood 
1 blue engine cover   underhood 
2 spark plug          underhood 
3 rear panel          bodywork 
4 black rear panel    bodywork 
5 blue engine         underhood 

最接近的是：

d=catg_df.apply('|'.join).to_dict()

desc_df['Category'] = desc_df['col1'].apply(lambda x : ''.join([z if pd.Series(x).str.contains(y).values else '' for z,y in d.items()]))

但我最终在字符串中找到了“引擎”和“引擎盖”： 描述

col1                   Category
0 engine              underhood 
1 blue engine cover   bodyworkunderhood 
2 spark plug          underhood 
3 rear panel          bodywork 
4 black rear panel    bodywork 
5 blue engine         underhood 

有什么方法我可以使用，也许如果它找到“引擎盖”然后分类使用这个类别，而不是移动到“引擎”。你知道吗

Answer 1

您可以通过迭代字典来解决此问题：

from collections import OrderedDict

d = OrderedDict([(k, '|'.join(catg_df[k].tolist())) for k in catg_df.columns[::-1]])

for k, v in d.items():
    desc_df.loc[desc_df['col1'].str.contains(v), 'Category'] = k

结果

print(desc_df)

                col1   Category
0             engine  underhood
1  blue engine cover   bodywork
2         spark plug  underhood
3         rear panel   bodywork
4   black rear panel   bodywork
5        blue engine  underhood

解释

• 对于字典中的每一项，检查str.contains条件与regex值，并将键分配给“Category”列。你知道吗
• 使用collections.OrderedDict为列赋予优先级。你知道吗
• 在这种情况下，可以在d的构造中反转列的迭代顺序。你知道吗

Answer 2

一种方法是使用difflib获取最接近的值和lambda：

首先创建映射器：

from difflib import get_close_matches
mapper = {val:k for k, v in catg_df.to_dict('list').items() for val in v}
print(mapper)

所以，mapper应该是：

{'Windscreen': 'Glass',
 'demister': 'Glass',
 'engine': 'underhood',
 'engine cover': 'bodywork',
 'oil filter': 'underhood',
 'rear panel': 'bodywork',
 'side panel': 'bodywork',
 'spark plug': 'underhood',
 'window': 'Glass'}

现在，使用lambda和difflib来查找最接近的值：

# avoid calling mapper.keys() in lambda 
keys = mapper.keys()
desc_df['Category'] = desc_df['col1'].apply(lambda row: mapper[get_close_matches(row, keys)[0]])

结果：

                col1   Category
0             engine  underhood
1  blue engine cover   bodywork
2         spark plug  underhood
3         rear panel   bodywork
4   black rear panel   bodywork
5        blue engine  underhood