Question

我需要帮助对看不见的数据进行分类。我有一组样本数据。

    ID         Comment                Category
    2017_01    inadequate stock       Availability
    2017_02    Too many failures      Quality
    2017_03    no documentation       Customer Service
    2017_04    good product           Satisfied
    2017_05    long delivery times    Delivery

我使用这些数据训练了一个多级文本分类器。我使用MultinomialNB和SVM测试了数据的拟合，我选择了SVM作为最终模型

# Support Vector Machines - calculating the SVM Fit
from sklearn.linear_model import SGDClassifier
text_clf_svm = Pipeline([('vect', CountVectorizer()), ('tfidf', 
TfidfTransformer()),('clf-svm', SGDClassifier(loss='hinge', 
penalty='l2',alpha=1e-3, n_iter=5, random_state=42))])

text_clf_svm = text_clf_svm.fit(X_train, y_train)
predicted_svm = text_clf_svm.predict(X_train)
np.mean(predicted_svm == y_train)
0.8850102669404517

我在今年的评论中对模型进行了测试

print(text_clf_svm.predict(["This is obsolete and being replaced by another product. not very robust and we have had many failures"]))

[＆＃39;质量＆＃39;]

问题：如何从2018年（下方）传递看不见的数据，以便按上述方式进行分类？

   ID          Comment                            Category
   2018_01     This product is obsolete 
   2018_02     Tech Support takes too long  
   2018_03     2 out of 3 products failed   
   2018_04     Delivery to APAC takes too long

Answer 1

我刚刚想出了解决方案。如果我想将预测结果映射回原始数据框，我必须将新列添加到旧数据框

#Bring in new data and predict
import pandas as pd
df_p = pd.read_csv("comment_to_predict.csv", sep = ',', usecols = range(2), encoding='iso-8859-1')

#Map the predictions to the original data frame (df_p in my case)
df_p['category'] = text_clf_svm.predict(df_p['comment'])

从训练有素的分类器python #Scikitlearn #SVM做出预测

1 个答案: