Question

#Load libraries

import numpy as np

from sklearn.naive_bayes import MultinomialNB

from sklearn.feature_extraction.text import CountVectorizer

from sklearn.feature_extraction.text import TfidfVectorizer 
#Create text
text_data = np.array(['Tim is smart!',
                      'Joy is the best',
                      'Lisa is dumb',
                      'Fred is lazy',
                      'Lisa is lazy'])
#Create target vector

y = np.array([1,1,0,0,0])

#feature extraction

tfidf_vectorizer = TfidfVectorizer()
X = tfidf_vectorizer.fit_transform(text_data)


mnb = MultinomialNB(alpha = 1, fit_prior = True, class_prior = None)

mnb.fit(X,y)

print(tfidf_vectorizer.get_feature_names())

#output

['best', 'dumb', 'fred', 'is', 'joy', 'lazy', 'lisa', 'smart', 'the', 'tim']

print(mnb.class_log_prior_)

#output

[-0.51082562 -0.91629073]

print(mnb.feature_log_prob_) 

#output

[[-2.70956132 -2.16160272 -2.16160272 -1.97680838 -2.70956132 -1.90253959
  -1.90253959 -2.70956132 -2.70956132 -2.70956132]

 [-2.1671355  -2.60968347 -2.60968347 -2.14937805 -2.1671355  -2.60968347
  -2.60968347 -2.09680488 -2.1671355  -2.09680488]]
  
  
X_log_proba = mnb.predict_log_proba(tfidf_vectorizer.transform(["best"]))
print(mnb.predict(tfidf_vectorizer.transform(text_)))

#output
[1]

print(X_log_proba)  
#output

[[-0.76397049 -0.62700977]]

我正在尝试计算 X_log_proba 的值。据我所知，这个值来自日志先验类+类中特征的日志。但例如，要预测仅包含 -2.167 的“最佳”单词的文本，它会给出 [[-0.76397049 -0.62700977]]。

但是当我尝试将特征“最佳”（-2.167）的对数概率添加到对数先验类（-0.91629073）时，结果不是-0.627。有人能解释一下模型如何通过给定文本在此示例中仅包含单词“best”来给出 -0.627 的预测对数概率？

如何计算预测对数概率朴素贝叶斯？

0 个答案: