Error in result

[Dhruval59]

Hi there, I trained the model but the results seem quite adverse. The predicted value is far away from the target value.

cv_id = 24861.0 target = 0.0 predicted = 1.0 probability = 0.592 cv_id = 25990.0 target = 1.0 predicted = 0.0 probability = 0.450 cv_id = 13568.0 target = 0.0 predicted = 1.0 probability = 0.600 cv_id = 27752.0 target = 0.0 predicted = 1.0 probability = 0.901 cv_id = 25574.0 target = 0.0 predicted = 1.0 probability = 0.930 cv_id = 17869.0 target = 0.0 predicted = 1.0 probability = 0.668 cv_id = 22643.0 target = 1.0 predicted = 0.0 probability = 0.056 cv_id = 25982.0 target = 0.0 predicted = 1.0 probability = 0.982 cv_id = 17962.0 target = 1.0 predicted = 0.0 probability = 0.099 cv_id = 14567.0 target = 1.0 predicted = 0.0 probability = 0.090 cv_id = 24423.0 target = 1.0 predicted = 0.0 probability = 0.488 cv_id = 24036.0 target = 0.0 predicted = 1.0 probability = 0.965 cv_id = 15627.0 target = 0.0 predicted = 1.0 probability = 0.988 cv_id = 19418.0 target = 0.0 predicted = 1.0 probability = 0.988

It would be great if you check this and suggest a way.

[florex]

Hi, That is very strange. There is no ligne where the target value matches the predicted value. Even a random model would have done better in this case.

The method bas been tested and its works quiet well. Please, revise the code and check if the values of variables are correct.

I will try to run the code again by my side...

Le vendredi 11 novembre 2022, Dhruval Sharma @.***> a écrit :

Hi there, I trained the model but the results seem quite adverse. The predicted value is far away from the target value.

cv_id = 24861.0 target = 0.0 predicted = 1.0 probability = 0.592 cv_id = 25990.0 target = 1.0 predicted = 0.0 probability = 0.450 cv_id = 13568.0 target = 0.0 predicted = 1.0 probability = 0.600 cv_id = 27752.0 target = 0.0 predicted = 1.0 probability = 0.901 cv_id = 25574.0 target = 0.0 predicted = 1.0 probability = 0.930 cv_id = 17869.0 target = 0.0 predicted = 1.0 probability = 0.668 cv_id = 22643.0 target = 1.0 predicted = 0.0 probability = 0.056 cv_id = 25982.0 target = 0.0 predicted = 1.0 probability = 0.982 cv_id = 17962.0 target = 1.0 predicted = 0.0 probability = 0.099 cv_id = 14567.0 target = 1.0 predicted = 0.0 probability = 0.090 cv_id = 24423.0 target = 1.0 predicted = 0.0 probability = 0.488 cv_id = 24036.0 target = 0.0 predicted = 1.0 probability = 0.965 cv_id = 15627.0 target = 0.0 predicted = 1.0 probability = 0.988 cv_id = 19418.0 target = 0.0 predicted = 1.0 probability = 0.988

It would be great if you check this and suggest a way.

— Reply to this email directly, view it on GitHub https://github.com/florex/preprocessing/issues/4, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB7HQSVYDYO2JNYYW7WAQOTWHXTHVANCNFSM6AAAAAAR5JTTGA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

-- Florentin Jiechieu Phd student, M.Sc, M.Eng, ITIL Certified Publications : https://rdcu.be/b6xFt https://arima.episciences.org/5802/pdf https://arxiv.org/abs/2010.03724 https://arxiv.org/abs/2010.03724 Stackoverflow : https://stackoverflow.com/users/3760878/florex Github : https://github.com/florex Play Store APP (under testing) : https://play.google.com/store/apps/details?id=com.jieffy.quizstorm Zone contenant les pièces jointes

[Dhruval59]

The Model predicts the following output after running multi_cnn.py: recall=0.9786 precision=0.8514 accuracy=0.8358

However, the above results suggest otherwise while running cnn_classifier.py. The output dir has a size of 20 GB after converting to matrices by preprocessing.

Thank you for trying it out again. Waiting for the results after your run.

[florex]

I understand now why you have these results and there is no problem with that.

Indeed, from line 133 to 137 (cnn_w2vec.py), I print only incorrect predictions. You will notice that at line 134, I check if the value predicted is different from the target value before printing the line.

Doing that, I wanted to recover resumes where the model failed at predicting the target class, go through the resume and figure out why the model failed at predicting the target value.

Le sam. 12 nov. 2022 à 20:56, Dhruval Sharma @.***> a écrit :

The Model predicts the following output after running multi_cnn.py: recall=0.9786 precision=0.8514 accuracy=0.8358

However, the above results suggest otherwise while running cnn_classifier.py. The output dir has a size of 20 GB after converting to matrices by preprocessing.

Thank you for trying it out again. Waiting for the results after your run.

— Reply to this email directly, view it on GitHub https://github.com/florex/preprocessing/issues/4#issuecomment-1312560198, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB7HQSRU5IPEAW2T65S25Y3WH7Y6ZANCNFSM6AAAAAAR5JTTGA . You are receiving this because you commented.Message ID: @.***>

-- Florentin Jiechieu https://play.google.com/store/apps/details?id=com.jieffy.quizstorm

[Dhruval59]

Yes. So need to change that condition in order to get the same result as the target values. I tried and worked well on test set.

cv_id = 10637.0 target = 0.0 predicted = 0.0 probability = 0.000 cv_id = 18087.0 target = 1.0 predicted = 1.0 probability = 0.997 cv_id = 12886.0 target = 1.0 predicted = 1.0 probability = 0.995 cv_id = 21953.0 target = 0.0 predicted = 0.0 probability = 0.037 cv_id = 26236.0 target = 1.0 predicted = 1.0 probability = 0.998 cv_id = 9195.0 target = 1.0 predicted = 1.0 probability = 0.989 cv_id = 22337.0 target = 0.0 predicted = 0.0 probability = 0.000 cv_id = 8283.0 target = 1.0 predicted = 1.0 probability = 0.978 cv_id = 6690.0 target = 1.0 predicted = 1.0 probability = 0.998 cv_id = 19072.0 target = 0.0 predicted = 0.0 probability = 0.007 cv_id = 15846.0 target = 0.0 predicted = 0.0 probability = 0.001 cv_id = 13744.0 target = 1.0 predicted = 1.0 probability = 0.988 cv_id = 27350.0 target = 0.0 predicted = 0.0 probability = 0.001 cv_id = 25112.0 target = 1.0 predicted = 1.0 probability = 0.998 cv_id = 9402.0 target = 1.0 predicted = 1.0 probability = 0.975