Transforming Dataframe for logistic regression (one hot encoding)

Question

I'm trying to run a logistic regression in python. My data consists of both numeric and categorical data. I would like to use gender, age and food preference to predict if someone likes cats.

I'm thinking that I would need to do one hot encoding on Food_preference (see below) but not sure exactly how to do it. Could you please help? thanks!

Original dataframe

Name    Gender  Age Like_cats   Food_preference
John    M   30  Yes Apple
John    M   30  Yes Orange
John    M   30  Yes Steak
Amy F   20  No  Apple
Amy F   20  No  Grape

Desired dataframe

Name    Gender  Age Like_cats   Apple   Orange  Steak   Grape
John    M   30  Yes 1   1   1   0
Amy F   20  No  1   0   0   1

Answer 1

You can use LabelEncoder to transform your string features to numeric features.

https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.LabelEncoder.html#sklearn.preprocessing.LabelEncoder

Here's a working code with same data structure as yours:

from sklearn.linear_model import LogisticRegression
import pandas as pd
from sklearn import preprocessing
import numpy as np



X = pd.DataFrame([['a', 0], ['b', 1], ['a', 5], ['b', 100]])
y = [0, 1, 0, 1]

X_n = [[]]*len(X.columns)

i = 0
for c in X.columns:
    if type(X[c].iloc[0]) == str: # if features are string encode them
        le = preprocessing.LabelEncoder()
        le.fit( list(set(X[c])) )
        X_n[i] = le.transform(X[c]) 
    else: # already numeric features
        X_n[i] = list(X[c])
    i += 1

X_n = np.array(X_n).T # transposing to make rows as per sample feature
print(X_n)

clf = LogisticRegression(random_state=0).fit(X_n, y)