[Dimension Reduction] LDA 변환

 

 

LDA¶

 

In [1]:

from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.preprocessing import StandardScaler
from sklearn.datasets import load_iris

import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline

 

In [2]:

iris = load_iris()
iris_scaled = StandardScaler().fit_transform(iris.data)

 

 

• LDA 변환

 

In [3]:

lda = LinearDiscriminantAnalysis(n_components=2)
lda.fit(iris_scaled, iris.target)
iris_lda = lda.transform(iris_scaled)
print(iris_lda.shape)

 

 

(150, 2)

 

 

• LDA 변환된 입력 데이터 값 시각화

 

In [4]:

lda_columns=['lda_component_1','lda_component_2']
iris_lda_df = pd.DataFrame(iris_lda,columns=lda_columns)
iris_lda_df['target']=iris.target

#setosa는 세모, versicolor는 네모, virginica는 동그라미로 표현
markers=['^', 's', 'o']

#setosa의 target 값은 0, versicolor는 1, virginica는 2. 각 target 별로 다른 shape으로 scatter plot
for i, marker in enumerate(markers):
    x_axis_data = iris_lda_df[iris_lda_df['target']==i]['lda_component_1']
    y_axis_data = iris_lda_df[iris_lda_df['target']==i]['lda_component_2']

    plt.scatter(x_axis_data, y_axis_data, marker=marker,label=iris.target_names[i])

plt.legend(loc='upper right')
plt.xlabel('lda_component_1')
plt.ylabel('lda_component_2')
plt.show()