We can use Linear Reg. for binary classification but it can give y as more than 1 or less than 0 even when possible classes are 1 and 0. When we use Linear Reg. for binary classification it gives same result as LDA.

NOTE: Put Logistic Function equatiom and cost function equation here:

We can use logistic regression for multi-class classification but not used that often, it is mostly used for binary classification.

Func Imports¶

In [1]:

import pandas as pd

from sklearn import datasets
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score, classification_report

Data Prep¶

In [2]:

# using sklearn default dataset
iris = datasets.load_iris()
iris.keys()

Out[2]:

dict_keys(['data', 'target', 'target_names', 'DESCR', 'feature_names'])

In [3]:

# Prepare data for training and testing
X = pd.DataFrame(iris['data'], columns=iris['feature_names'])
y = pd.Series(iris['target'], name='species')\
      .apply(lambda x: iris['target_names'][x])

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.5, random_state=42)

Model Fitting¶

In [4]:

# Initialize the model
model = LogisticRegression(multi_class='multinomial', solver='newton-cg')

# Model Fitting
model.fit(X_train, y_train)

# Predictions
y_predict = model.predict(X_test)

Performance Check¶

In [5]:

accuracy_score(y_test, y_predict)

Out[5]:

1.0

In [6]:

print(classification_report(y_test, y_predict))

             precision    recall  f1-score   support

     setosa       1.00      1.00      1.00        29
 versicolor       1.00      1.00      1.00        23
  virginica       1.00      1.00      1.00        23

avg / total       1.00      1.00      1.00        75