データの正規化と予測精度|データの前処理

データの正規化と予測精度

使用するデータセットは、cancerです。

データの正規化データの前処理という記事もあります。

正規化なし

from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.datasets import load_breast_cancer

cancer = load_breast_cancer()
train_X, val_X, train_y, val_y = train_test_split(cancer.data, cancer.target, test_size=0.3, random_state=0)

svm = SVC(C=100)
svm.fit(train_X, train_y)
pred = svm.predict(val_X)
print(accuracy_score(val_y, pred))

平均0分散1の正規化

$$
\frac{X – \bar{X}}{\sigma}
$$

from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.datasets import load_breast_cancer
from sklearn.preprocessing import StandardScaler


cancer = load_breast_cancer()
train_X, val_X, train_y, val_y = train_test_split(cancer.data, cancer.target, test_size=0.3, random_state=0)


scaler_std = StandardScaler()
scaler_std.fit(train_X)

std_train_X = scaler_std.transform(train_X)
std_val_X = scaler_std.transform(val_X)

svm = SVC(C=100)
svm.fit(std_train_X, train_y)
pred = svm.predict(s_val_X)
print(accuracy_score(val_y, pred))

最小0最大1の正規化

$$
\frac{X – Min(X)}{Max(X) – Min(X)}
$$

from sklearn.svm import SVC
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score
from sklearn.datasets import load_breast_cancer
from sklearn.preprocessing import MinMaxScaler

cancer = load_breast_cancer()
train_X, val_X, train_y, val_y = train_test_split(cancer.data, cancer.target, test_size=0.3, random_state=0)

#標準化
scaler_mm = MinMaxScaler()
scaler_mm.fit(train_X)
mm_train_X = scaler_mm.transform(train_X)
mm_val_X = scaler_mm.transform(val_X)

svm = SVC(C=100)
svm.fit(mm_train_X, train_y)
pred = svm.predict(mm_val_X)
print(accuracy_score(val_y, pred))

予測精度

まとめ

正規化をすることで、予測精度がよくなることが確認できた。

今回のデータセットでは、平均0分散1の正規化より、最小0最大1の正規化の方が、予測精度が良かった。

参考