I'm trying to create dummy variables for a variable that has text data in rows.

Data in 1st row is:
{"Wireless Internet","Air conditioning",Kitchen,Heating,"Family/kid friendly",Essentials,"Hair dryer",Iron,"translation missing: en.hosting_amenity_50"}

and Data in 2nd row is:
{TV,"Cable TV",Internet,"Wireless Internet",Kitchen,"Indoor fireplace","Buzzer/wireless intercom",Heating,Washer,Dryer,"Smoke detector","Carbon monoxide detector","First aid kit","Fire extinguisher",Essentials} and many more.

What I now want to do is, to create dummy variables out of that variable. For example from the above data:
one variable named Wireless Internet with 0 ans 1 in rows &
another variable named Cable TV with 0 and 1 in rows &
another variable named Kitchen with 0 and 1 in rows and so on.

sklearn for python has OneHotEncoder class which creates dummy variable named everything in a row considering all rows with unique values. That is not what I want to do here. I first have to split text in all rows and create dummy variables for them. How do I do that?

Expected results are, multiple columns like
Wireless Internet Cable TV Kitchen
1 0 1
0 1 1
1 0 1

example of transformation link to data(column named amenities) - https://www.kaggle.com/stevezhenghp/airbnb-price-prediction


1 Answer 1


On the kaggle page, there is a kernel available which focuses on exactly your problem. This is the code from user JAGADEESHWARA VARA PRASAD:

# load your dataset into df
df = pd.read_csv("../input/train.csv")

# trimm { } symbols and split at ',' and trimm " from each word
l = [[word.strip('[" ]') for word in row[1:-1].split(',')]
     for row in list(df['amenities'])]

# form a set of distinct text
cols = set(word for row in l for word in row)

# create and fill new data frame
new_df = pd.DataFrame(columns=cols)
for row_idx in range(len(l)):
    for col in cols:
        new_df.loc[row_idx,col]=int(col in l[row_idx])


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.