Sklearn SVM - 如何获得错误预测的列表？

Question

我不是专家用户。 我知道我可以获得混淆矩阵，但我想获取以错误方式分类的行的列表，以便在分类后研究它们。

在 stackoverflow 上，我发现这个Can I get a list of wrong predictions in SVM score function in scikit-learn但我不确定是否理解了所有内容。

这是一个示例代码。

# importing necessary libraries
from sklearn import datasets
from sklearn.metrics import confusion_matrix
from sklearn.model_selection import train_test_split

# loading the iris dataset
iris = datasets.load_iris()

# X -> features, y -> label
X = iris.data
y = iris.target

# dividing X, y into train and test data
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state = 0)

# training a linear SVM classifier
from sklearn.svm import SVC
svm_model_linear = SVC(kernel = 'linear', C = 1).fit(X_train, y_train)
svm_predictions = svm_model_linear.predict(X_test)

# model accuracy for X_test  
accuracy = svm_model_linear.score(X_test, y_test)

# creating a confusion matrix
cm = confusion_matrix(y_test, svm_predictions)

要遍历行并找到错误的行，建议的解决方案是：

predictions = clf.predict(inputs)
for input, prediction, label in zip(inputs, predictions, labels):
  if prediction != label:
    print(input, 'has been classified as ', prediction, 'and should be ', label) 

我不明白什么是“输入”/“输入”。 如果我将此代码改编为我的代码，如下所示：

for input, prediction, label in zip (X_test, svm_predictions, y_test):
  if prediction != label:
    print(input, 'has been classified as ', prediction, 'and should be ', label)

我获得：

[6.  2.7 5.1 1.6] has been classified as  2 and should be  1

第 6 行是错误的行吗？ 6.后面的数字是什么？ 我问这个是因为我在比这个更大的数据集上使用相同的代码，所以我想确保我做的是正确的事情。 我没有发布其他数据集，因为不幸的是我不能，但问题是我获得了这样的东西：

  (0, 253)  0.5339655767137572
  (0, 601)  0.27665553856928027
  (0, 1107) 0.7989633757962163 has been classified as  7 and should be  3
  (0, 885)  0.3034934766501018
  (0, 1295) 0.6432561790864061
  (0, 1871) 0.7029318585026516 has been classified as  7 and should be  6
  (0, 1020) 1.0 has been classified as  3 and should be  8

当我计算最后输出的每一行时，我得到了测试集行的两倍......所以我不确定我是否正在分析错误的预测结果列表......

Answer 1

第 6 行是错误的行吗？ 6.后面的数字是什么？

不 - [6. 2.7 5.1 1.6] [6. 2.7 5.1 1.6]是实际样本（即其特征）。 要获取错误行的索引，我们应该稍微修改for循环：

for idx, input, prediction, label in zip(enumerate(X_test), X_test, svm_predictions, y_test):
    if prediction != label:
        print("No.", idx[0], 'input,',input, ', has been classified as', prediction, 'and should be', label) 

结果现在是

No. 37 input, [ 6.   2.7  5.1  1.6] , has been classified as 2 and should be 1

这意味着X_test[37] ，即[ 6. 2.7 5.1 1.6] ，已被我们的 SVM 预测为 2，而其真实标签为 1。

让我们确认这个读数：

X_test[37]
# array([ 6. ,  2.7,  5.1,  1.6])

svm_predictions[37]
# 2

y_test[37]
# 1

这个结果与你的混淆矩阵cm ，它确实显示了X_test只有一个错误分类的样本：

cm
# result:
array([[13,  0,  0],
       [ 0, 15,  1],
       [ 0,  0,  9]], dtype=int64)

一个更优雅的for循环，因为枚举包括样本本身，可以是：

for idx, prediction, label in zip(enumerate(X_test), svm_predictions, y_test):
    if prediction != label:
        print("Sample", idx, ', has been classified as', prediction, 'and should be', label) 

这使

Sample (37, array([ 6. ,  2.7,  5.1,  1.6])) , has been classified as 2 and should be 1

Answer 2

如果您只想获取错误分类实例的列表，您可以执行以下操作：

# with the following sentence you can get a mask of the items bad classified
mask = np.logical_not(np.equal(y_test, predictions))
# Now you can use the mask to see the elements bad classified:
print(f"Elements wrong classified: {X_test[mask]}")
print(f"Prediction by the model for each of those elements: {predictions[mask]}")
print(f"Actual value for each of those elements: {np.asarray(y_test)[mask]}")