CODEWITHSUNDEEP
pythonpandasperformanceoptimizationstring-matching
IdoKendo
IdoKendo

Reputation: 115

Match two data frames by substring in python

I have two big data frames (1000s of rows), and I need to match them by substring, for example:

df1:

Id    Title
1     The house of pump
2     Where is Andijan
3     The Joker
4     Good bars in Andijan
5     What a beautiful house

df2:

Keyword
house
andijan
joker

And the expected output is:

Id    Title                    Keyword
1     The house of pump        house
2     Where is Andijan         andijan
3     The Joker                joker
4     Good bars in Andijan     andijan
5     What a beautiful house   house

Now, I've sort of written a very non-efficient way to match it, but for the actual size of the data frames it runs for an extremely long time:

for keyword in df2.to_dict(orient='records'):
    df1['keyword'] = np.where(creative_df['title'].str.contains(keyword['keyword']), keyword['keyword'], df1['keyword'])

Now, I'm sure there's a more pandas-friendy and efficient way to do the same, and also have it run in a reasonable time.

Upvotes: 3

Views: 115

Answers (2)

rudolfovic
rudolfovic

Reputation: 3276

Developing @BENY's solution a little further in order to be able to get multiple keywords per title:

regex = '|'.join(keywords['Keyword'])
keywords = df['Title'].str.findall(regex, flags=re.IGNORECASE)
keywords_exploded = pd.DataFrame(keywords.explode().dropna())
df.merge(keywords_exploded, left_index=True, right_index=True)

Upvotes: 0

BENY
BENY

Reputation: 323326

Let us try findall

import re
df1['new'] = df1.Title.str.findall('|'.join(df2.Keyword.tolist()),flags= re.IGNORECASE).str[0]
df1
   Id                   Title      new
0   1       The house of pump    house
1   2        Where is Andijan  Andijan
2   3               The Joker    Joker
3   4    Good bars in Andijan  Andijan
4   5  What a beautiful house    house

Upvotes: 3

Related Questions