ValueError：无法将字符串转换为浮点数：sklearn、numpy、panda

Question

Im trying to convert car names from NumPy array to numeric values to use for linear regressor. The label encoder gives warning: ValueError: could not convert string to float: 'porsche' Can someone help, please?

Heres the code:

 from sklearn.preprocessing import StandardScaler
 from sklearn.preprocessing import LabelEncoder, OneHotEncoder
 enc = LabelEncoder()
 enc.fit_transform(Z[:,0:1])
 onehotencoder = OneHotEncoder(categorical_features = [0])
 Z = onehotencoder.fit_transform(Z).toarray()`

and outoput: ValueError: could not convert string to float: 'porsche'

And here is the array: Array name = Z, type str416,

Answer 1

For one hot encoding, I would suggest you to use pd.get_dummies instead, much easier to use:

# make sure Z is a dataframe
X = pd.get_dummies(Z).values

If you want to use sklearn's OHE, you can refer to the following example:

from sklearn.preprocessing import StandardScaler, LabelEncoder, OneHotEncoder

df = pd.DataFrame({'a':['audi','porsche','audi'], 'b':[1,2,3]})
ohe = OneHotEncoder()

mat = ohe.fit_transform(df[['a']])

# view the contents of array
mat.todense()

matrix([[1., 0.],
        [0., 1.],
        [1., 0.]])

# get feature names
ohe.get_feature_names()
array(['x0_audi', 'x0_porsche'], dtype=object)