How to map DataFrame values from strings to float in Python-CodePudding

I have a data frame I am trying to do predictions on house-prices based on the data such as square footage or if it has central air. I need to convert string values to numbers in order to model it. For example the values in the CentralAir column are 'N' or 'Y' which I want to be 0 and 1 respectively.

# pull data into target (y) and predictors (X) using other standard predictors; 
train_y2 = df_train.SalePrice

#convert strings to float so we can use predictors like the neighborhood and building type
central_air_mapping = {'N':0, 'Y':1}
df_train['CentralAir'] = df_train.map(central_air_mapping)

predictor_cols2 = ['CentralAir']

# Create training predictors data
train_X2 = df_train[predictor_cols2]

my_model2 = RandomForestRegressor()
my_model2.fit(train_X2, train_y2)

Then it returns

AttributeError: 'DataFrame' object has no attribute 'map'

CodePudding user response：

def ToNum(c):
    if c == "Y":
        return 1
    else:
        return 0

df_train["CentralAir"] = df_train["CentralAir"].apply(ToNum)

CodePudding user response：

.map() is defined for Series, not DataFrames. This is why you are getting an error.

central_air_mapping = {'N':0, 'Y':1}
df_train['CentralAir'] = df_train['CentralAir'].map(central_air_mapping)
                                      ^^^^
                                 this was missing