Scikit-learn - 在测量精确度时获取NAN值。

Question

我对样本数据分类正面和负面情绪。我使用了下面的代码片断。

一切看起来都OK，直到第20行打印预期的预测。

但是，当我尝试使用度量标准来衡量准确性时，它给了我“NAN”值。你可以请检阅我的代码，并帮我找出问题。

from sklearn.naive_bayes import MultinomialNB 
from sklearn.feature_extraction.text import TfidfVectorizer 
from sklearn import metrics 
import csv 

# Read in the training data. 
with open("/Users/max/train.csv", 'r') as file: 
    reviews = list(csv.reader(file)) 

with open("/Users/max/test.csv",'r') as file: 
    test_reviews = list(csv.reader(file)) 

vectorizer = TfidfVectorizer(min_df=1) 
train_features = vectorizer.fit_transform([review[0] for review in reviews]) 
test_features = vectorizer.transform([test_review[0] for test_review in test_reviews]) 

nb = MultinomialNB() 
nb.fit(train_features, [int(review[1]) for review in reviews]) 

predictions = nb.predict(test_features) 
print("prediction : {0}".format(predictions)) 

actual = [int(r[1]) for r in test_reviews] 
fpr, tpr, threshold = metrics.roc_curve(actual, predictions, pos_label=1) 
print("Multinomial naive bayes AUC: {0}".format(metrics.auc(fpr, tpr))) 

集样本以这种格式

i like google , 1 
i dont really like microsoft , -1 

这里是控制台输出你没有在你的数据的真正积极的实例

prediction : [1 -1] 
/Library/Python/2.7/site-packages/sklearn/metrics/ranking.py:496: UndefinedMetricWarning: No positive samples in y_true, true positive value should be meaningless UndefinedMetricWarning) 
Multinomial naive bayes AUC: nan 

Answer 1

概率。