Encoding two categorial data present in same dataset in Deep learning

Question

I have a dataset with columns reason and issue.

I wanted to encode it as:

enc = OneHotEncoder()
reason_no_enc = enc.fit_transform(temp['REASON NO'].values.reshape(-1, 1)).toarray()
issue_enc = enc.fit_transform(temp['Issue'].values.reshape(-1, 1)).toarray()

But I realized it is creating problem, the later one issue_enc is considered encoded, when I try to inverse reason_no_enc, it generates an error.

How to handle it?

Answer 1

You have to use different instances of OHE for each column like this:

# fit encoder using 'REASON NO' data
# later use this instance of OHE to decode 'REASON NO' data
ohe_reason = OneHotEncoder()
reason_no_enc = ohe_reason.fit_transform(temp['REASON NO'].values.reshape(-1, 1)).toarray()

# fit encoder using 'Issue' data
# later use this instance of OHE to decode 'Issue' data
ohe_issue = OneHotEncoder()
issue_enc = ohe_issue.fit_transform(temp['Issue'].values.reshape(-1, 1)).toarray()

And also you can use one instance of OHE for both categories like this:

enc = OneHotEncoder()
encoded_arr = enc.fit_transform(temp[['REASON NO', 'Issue']])