串聯多種特徵提取方法#

在許多現實世界的例子中，有很多方法可以從資料集中提取特徵。通常，結合多種方法以獲得良好的效能是有益的。此範例示範如何使用 FeatureUnion 來組合由 PCA 和單變數選擇獲得的特徵。

使用此轉換器組合特徵的好處是，它允許對整個過程進行交叉驗證和網格搜尋。

此範例中使用的組合對此資料集並非特別有用，僅用於說明 FeatureUnion 的用法。

Combined space has 3 features
Fitting 5 folds for each of 18 candidates, totalling 90 fits
[CV 1/5; 1/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV 1/5; 1/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 2/5; 1/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV 2/5; 1/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 3/5; 1/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV 3/5; 1/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1;, score=0.867 total time=   0.0s
[CV 4/5; 1/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV 4/5; 1/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 5/5; 1/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1
[CV 5/5; 1/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 2/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV 1/5; 2/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=1;, score=0.900 total time=   0.0s
[CV 2/5; 2/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV 2/5; 2/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 3/5; 2/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV 3/5; 2/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=1;, score=0.867 total time=   0.0s
[CV 4/5; 2/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV 4/5; 2/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=1;, score=0.933 total time=   0.0s
[CV 5/5; 2/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=1
[CV 5/5; 2/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 3/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=10
[CV 1/5; 3/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=10;, score=0.933 total time=   0.0s
[CV 2/5; 3/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=10
[CV 2/5; 3/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 3/5; 3/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=10
[CV 3/5; 3/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=10;, score=0.900 total time=   0.0s
[CV 4/5; 3/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=10
[CV 4/5; 3/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=10;, score=0.933 total time=   0.0s
[CV 5/5; 3/18] START features__pca__n_components=1, features__univ_select__k=1, svm__C=10
[CV 5/5; 3/18] END features__pca__n_components=1, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 1/5; 4/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1
[CV 1/5; 4/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 2/5; 4/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1
[CV 2/5; 4/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 3/5; 4/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1
[CV 3/5; 4/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 4/5; 4/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1
[CV 4/5; 4/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 5/5; 4/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1
[CV 5/5; 4/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 5/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=1
[CV 1/5; 5/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=1;, score=0.933 total time=   0.0s
[CV 2/5; 5/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=1
[CV 2/5; 5/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 3/5; 5/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=1
[CV 3/5; 5/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=1;, score=0.933 total time=   0.0s
[CV 4/5; 5/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=1
[CV 4/5; 5/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=1;, score=0.933 total time=   0.0s
[CV 5/5; 5/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=1
[CV 5/5; 5/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 6/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=10
[CV 1/5; 6/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=10;, score=0.967 total time=   0.0s
[CV 2/5; 6/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=10
[CV 2/5; 6/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=10;, score=0.967 total time=   0.0s
[CV 3/5; 6/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=10
[CV 3/5; 6/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=10;, score=0.933 total time=   0.0s
[CV 4/5; 6/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=10
[CV 4/5; 6/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=10;, score=0.933 total time=   0.0s
[CV 5/5; 6/18] START features__pca__n_components=1, features__univ_select__k=2, svm__C=10
[CV 5/5; 6/18] END features__pca__n_components=1, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
[CV 1/5; 7/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1
[CV 1/5; 7/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 2/5; 7/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1
[CV 2/5; 7/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 3/5; 7/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1
[CV 3/5; 7/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1;, score=0.867 total time=   0.0s
[CV 4/5; 7/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1
[CV 4/5; 7/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 5/5; 7/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1
[CV 5/5; 7/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 8/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=1
[CV 1/5; 8/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=1;, score=0.967 total time=   0.0s
[CV 2/5; 8/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=1
[CV 2/5; 8/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 3/5; 8/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=1
[CV 3/5; 8/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=1;, score=0.933 total time=   0.0s
[CV 4/5; 8/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=1
[CV 4/5; 8/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=1;, score=0.933 total time=   0.0s
[CV 5/5; 8/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=1
[CV 5/5; 8/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 9/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=10
[CV 1/5; 9/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=10;, score=0.967 total time=   0.0s
[CV 2/5; 9/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=10
[CV 2/5; 9/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=10;, score=0.967 total time=   0.0s
[CV 3/5; 9/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=10
[CV 3/5; 9/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=10;, score=0.900 total time=   0.0s
[CV 4/5; 9/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=10
[CV 4/5; 9/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=10;, score=0.933 total time=   0.0s
[CV 5/5; 9/18] START features__pca__n_components=2, features__univ_select__k=1, svm__C=10
[CV 5/5; 9/18] END features__pca__n_components=2, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 1/5; 10/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1
[CV 1/5; 10/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 2/5; 10/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1
[CV 2/5; 10/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 3/5; 10/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1
[CV 3/5; 10/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 4/5; 10/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1
[CV 4/5; 10/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 5/5; 10/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1
[CV 5/5; 10/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 11/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=1
[CV 1/5; 11/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 2/5; 11/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=1
[CV 2/5; 11/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=1;, score=1.000 total time=   0.0s
[CV 3/5; 11/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=1
[CV 3/5; 11/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=1;, score=0.933 total time=   0.0s
[CV 4/5; 11/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=1
[CV 4/5; 11/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 5/5; 11/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=1
[CV 5/5; 11/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 12/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=10
[CV 1/5; 12/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=10;, score=0.967 total time=   0.0s
[CV 2/5; 12/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=10
[CV 2/5; 12/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
[CV 3/5; 12/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=10
[CV 3/5; 12/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=10;, score=0.900 total time=   0.0s
[CV 4/5; 12/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=10
[CV 4/5; 12/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=10;, score=0.933 total time=   0.0s
[CV 5/5; 12/18] START features__pca__n_components=2, features__univ_select__k=2, svm__C=10
[CV 5/5; 12/18] END features__pca__n_components=2, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
[CV 1/5; 13/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1
[CV 1/5; 13/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 2/5; 13/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1
[CV 2/5; 13/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 3/5; 13/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1
[CV 3/5; 13/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 4/5; 13/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1
[CV 4/5; 13/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 5/5; 13/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1
[CV 5/5; 13/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 14/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=1
[CV 1/5; 14/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=1;, score=0.967 total time=   0.0s
[CV 2/5; 14/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=1
[CV 2/5; 14/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 3/5; 14/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=1
[CV 3/5; 14/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=1;, score=0.933 total time=   0.0s
[CV 4/5; 14/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=1
[CV 4/5; 14/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=1;, score=0.967 total time=   0.0s
[CV 5/5; 14/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=1
[CV 5/5; 14/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 15/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=10
[CV 1/5; 15/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 2/5; 15/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=10
[CV 2/5; 15/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 3/5; 15/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=10
[CV 3/5; 15/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=10;, score=0.933 total time=   0.0s
[CV 4/5; 15/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=10
[CV 4/5; 15/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=10;, score=0.967 total time=   0.0s
[CV 5/5; 15/18] START features__pca__n_components=3, features__univ_select__k=1, svm__C=10
[CV 5/5; 15/18] END features__pca__n_components=3, features__univ_select__k=1, svm__C=10;, score=1.000 total time=   0.0s
[CV 1/5; 16/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1
[CV 1/5; 16/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 2/5; 16/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1
[CV 2/5; 16/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 3/5; 16/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1
[CV 3/5; 16/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1;, score=0.933 total time=   0.0s
[CV 4/5; 16/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1
[CV 4/5; 16/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1;, score=0.967 total time=   0.0s
[CV 5/5; 16/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1
[CV 5/5; 16/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=0.1;, score=1.000 total time=   0.0s
[CV 1/5; 17/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=1
[CV 1/5; 17/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 2/5; 17/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=1
[CV 2/5; 17/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=1;, score=1.000 total time=   0.0s
[CV 3/5; 17/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=1
[CV 3/5; 17/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 4/5; 17/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=1
[CV 4/5; 17/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=1;, score=0.967 total time=   0.0s
[CV 5/5; 17/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=1
[CV 5/5; 17/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=1;, score=1.000 total time=   0.0s
[CV 1/5; 18/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=10
[CV 1/5; 18/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
[CV 2/5; 18/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=10
[CV 2/5; 18/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
[CV 3/5; 18/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=10
[CV 3/5; 18/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=10;, score=0.900 total time=   0.0s
[CV 4/5; 18/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=10
[CV 4/5; 18/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=10;, score=0.967 total time=   0.0s
[CV 5/5; 18/18] START features__pca__n_components=3, features__univ_select__k=2, svm__C=10
[CV 5/5; 18/18] END features__pca__n_components=3, features__univ_select__k=2, svm__C=10;, score=1.000 total time=   0.0s
Pipeline(steps=[('features',
                 FeatureUnion(transformer_list=[('pca', PCA(n_components=3)),
                                                ('univ_select',
                                                 SelectKBest(k=1))])),
                ('svm', SVC(C=10, kernel='linear'))])

# Authors: The scikit-learn developers
# SPDX-License-Identifier: BSD-3-Clause

from sklearn.datasets import load_iris
from sklearn.decomposition import PCA
from sklearn.feature_selection import SelectKBest
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import FeatureUnion, Pipeline
from sklearn.svm import SVC

iris = load_iris()

X, y = iris.data, iris.target

# This dataset is way too high-dimensional. Better do PCA:
pca = PCA(n_components=2)

# Maybe some original features were good, too?
selection = SelectKBest(k=1)

# Build estimator from PCA and Univariate selection:

combined_features = FeatureUnion([("pca", pca), ("univ_select", selection)])

# Use combined features to transform dataset:
X_features = combined_features.fit(X, y).transform(X)
print("Combined space has", X_features.shape[1], "features")

svm = SVC(kernel="linear")

# Do grid search over k, n_components and C:

pipeline = Pipeline([("features", combined_features), ("svm", svm)])

param_grid = dict(
    features__pca__n_components=[1, 2, 3],
    features__univ_select__k=[1, 2],
    svm__C=[0.1, 1, 10],
)

grid_search = GridSearchCV(pipeline, param_grid=param_grid, verbose=10)
grid_search.fit(X, y)
print(grid_search.best_estimator_)