如何在多次预测后计算误课率列？

Question

我试图找到一种方法来计算错误分类率后，通过运行几个预测迭代如下。我已尝试编写剩余的代码，但仍然无法运行。我做错了什么

predictions = df.copy()
y = df['gt']
noiter = 10
hits = 0
tpred = 0

for i in range(noiter):
    Xtrain, Xtest, ytrain, ytest = train_test_split(df,test_size=0.3,random_state=noiter)


    model = xgb.XGBClassifier()
    model.fit(X_train,y_train)
    pred_i = model.predict(X_test)
    newcol = 'npred_' + str(noiter)
    pred.loc[test.index,newcol] = pred_i

#now to calculate the misclassification rate
    if pred_i != 'NaN':
        tpred = tpred + 1


    if pred_i == test['gt']:
        hits = hits + 1

pred['missclassrate'] = hits/tpred

Answer 1

您可以获得混淆矩阵以确定误报和漏报：

from sklearn.metrics import confusion_matrix
print confusion_matrix(ytest, pred_i)