I am trying to first impute a categorical feature(embark from titanic dataset) using SimpleImputer and then do OHE encoding on it. The problem is, the imputer is creating a column of the original values that are in string. I tried setting False to copy parameter but it’s still creating a column of the original values. If I try to use this in pipeline to apply logistic regression, the string values breaks the code. How can avoid making the copy of the original values?

   numerical_transformer = ColumnTransformer([
      #impute missing values
      ('impute_age',SimpleImputer(copy=False),[0]),
      #scale the age and fare column
      ('scale_age',StandardScaler(),[0]),
      ('scale_fare',StandardScaler(),[1])
    ],remainder='passthrough')

    categorical_transformer = ColumnTransformer([
      #impute missing values
      ('impute_embark',SimpleImputer(strategy='most_frequent',copy=False),[3]),
      #encode embarked column
      ('ohe_embarked',OneHotEncoder(sparse_output=False,handle_unknown='ignore',drop='first'),[3]),
    ],remainder='passthrough')

    log_reg_transformer = LogisticRegression()

    pipe = make_pipeline(numerical_transformer,categorical_transformer,log_reg_transformer)
    pipe.fit(X_train,y_train)

The error I am getting is:
ValueError: could not convert string to float: ‘S’

Thanks in advance.

I tried setting copy parameter to false so that SimpleImputer doesn’t create the copy of original value