تحسين نموذج تصنيف النصوص باستخدام بايز الساذج - كتاب الذكاء الإصطناعي - الصف 12 - الفصل 1 - المملكة العربية السعودية

يمكن الآن إدخال أداة التمثيل بالمتجهات في مصنف بايز الساذج لبناء خط أنابيب نموذج تنبؤ جديد وتطبيقه على بيانات اختبار IMDb:

# train a new Naive Bayes Classifier on the newly vectorized data. model_tf =MultinomialNB() model_tf.fit(X_train_v2, Y_train) # create a new prediction pipeline. prediction_pipeline_tf = make_pipeline(vectorizer_tf, model_tf) # get predictions using the new pipeline. predictions_tf = prediction_pipeline_tf.predict(X_test_text_annotated) # print the achieved accuracy. accuracy_score(Y_test, predictions_tf)

0.8858

يحقق خط الأنابيب الجديد دقة تصل إلى 88.58%، وهو تحسن كبير بالمقارنة مع الدقة السابقة التي وصلت إلى 84.68%. يمكن الآن استخدام النموذج المحسن لإعادة النظر في مثال الاختبار الذي تم تصنيفه بشكل خاطئ بواسطة النموذج الأول:

# get the review example that confused the previous algorithm mistake_example_annotated=X_test_text_annotated[4600] print('\nReview:', mistake_example_annotated) # get the correct labels of this example. print('\nCorrect Label:', class_names[Y_test[4600]]) # get the prediction probabilities for this example. print('\nPrediction Probabilities for neg, pos:',prediction_pipeline_tf.predict_proba([mistake_example_annotated]))

Review: i_personally thought the movie was_pretty good very_good acting by tadanobu_ asano of ichi_the_killer fame i really can_t say much about the story but there_were parts that confused me a little_too much and overall i_thought the movie was just too lengthy other_than that however the movie contained superb_acting great fightings and a lot of the locations were beautifully_shot great effects and a lot of sword play another solid effort by tadanobu_asano in my_opinion well i really can_t say anymore about the movie but if_you re only outlook on asian_cinema is crouching_tiger hidden_ dragon or house of flying_daggers i_would suggest_you trying to rent_it but if_you re a die_hard asian_cinema fan i_would say this has to be in your_collection very_good japanese film

Correct Label: pos

Prediction Probabilities for neg, pos: [[0.32116538 0.67883462]]

150

وزارة التعليم Ministry of Education 2023 - 1447

يمكن الآن إدخال أداة التمثيل بالمتجهات في مصنف بايز الساذج لبناء خط أنابيب نموذج تنبؤ جديد وتطبيقه على بيانات اختبار IMDb:# train a new Naive Bayes Classifier on the newly vectorized data. model_tf =MultinomialNB() model_tf.fit(X_train_v2, Y_train)# create a new prediction pipeline. prediction_pipeline_tf = make_pipeline(vectorizer_tf, model_tf)# get predictions using the new pipeline. predictions_tf = prediction_pipeline_tf.predict(X_test_text_annotated)# print the achieved accuracy. accuracy_score(Y_test, predictions_tf)0.8858يحقق خط الأنابيب الجديد دقة تصل إلى 88.58%، وهو تحسن كبير بالمقارنة مع الدقة السابقة التي وصلت إلى 84.68%. يمكن الآن استخدام النموذج المحسن لإعادة النظر في مثال الاختبار الذي تم تصنيفه بشكل خاطئ بواسطة النموذج الأول:# get the review example that confused the previous algorithm mistake_example_annotated=X_test_text_annotated[4600] print('\nReview:', mistake_example_annotated)# get the correct labels of this example. print('\nCorrect Label:', class_names[Y_test[4600]])# get the prediction probabilities for this example. print('\nPrediction Probabilities for neg, pos:',prediction_pipeline_tf.predict_proba([mistake_example_annotated]))Review: i_personally thought the movie was_pretty good very_good acting by tadanobu_ asano of ichi_the_killer fame i really can_t say much about the story but there_were parts that confused me a little_too much and overall i_thought the movie was just too lengthy other_than that however the movie contained superb_acting great fightings and a lot of the locations were beautifully_shot great effects and a lot of sword play another solid effort by tadanobu_asano in my_opinion well i really can_t say anymore about the movie but if_you re only outlook on asian_cinema is crouching_tiger hidden_ dragon or house of flying_daggers i_would suggest_you trying to rent_it but if_you re a die_hard asian_cinema fan i_would say this has to be in your_collection very_good japanese film Correct Label: pos Prediction Probabilities for neg, pos: [[0.32116538 0.67883462]]2023 - 1447