ICACT20210116 Question.1
Questioner: James_Lan@htx.gov.sg 2021-02-09 오전 11:38:39 |
ICACT20210116 Answer.1
Answer by Auhor shenhanji@cnic.cn 2021-02-09 오전 11:38:39
Chrome  Click!! |
| Hi, thanks for the sharing. Can you share with us the ratio of training set, validation set and testing set, and the total number dataset used in the machine learning training? Thanks. |
We use the deployment website of China Science and technology network as the data set. The proportion of training set, verification set and test set is 9:2:2, and the total number of data sets is more than 3300 websites. The verification set and test set use the same data. |
ICACT20210116 Question.2
Questioner: dykim6@gmail.com 2021-02-09 오전 11:41:23 |
ICACT20210116 Answer.2
Answer by Auhor shenhanji@cnic.cn 2021-02-09 오전 11:41:23
Chrome Click!! |
| What is meant by 'recall'? |
Recall means the ratio of experimental results to labeled data. |
ICACT20210116 Question.3
Questioner: James_Lan@htx.gov.sg 2021-02-09 오전 11:39:16 |
ICACT20210116 Answer.3
Answer by Auhor shenhanji@cnic.cn 2021-02-09 오전 11:39:16
Chrome Click!! |
| Hi, thanks for the sharing.
Can you share with us the ratio of training set, validation set and testing set, and the total number dataset used in the machine learning training?
Thanks. |
We use the deployment website of China Science and technology network as the data set. The proportion of training set, verification set and test set is 9:2:2, and the total number of data sets is more than 3300 websites. The verification set and test set use the same data. |
ICACT20210116 Question.4
Questioner: Yiile@outlook.com 2021-02-10 오후 8:46:55 |
ICACT20210116 Answer.4
Answer by Auhor shenhanji@cnic.cn 2021-02-10 오후 8:46:55
Chrome Click!! |
| Dear author. How effective is the test?Is there an actual measurement of accuracy? |
The test is more effective, we choose the optimal parameters, the accuracy of the measurement data is based on our existing annotated web pages, with the enrichment of the data set, we can do better. |
ICACT20210116 Question.5
Questioner: hjcho1@korea.ac.kr 2021-02-10 오후 8:41:46 |
ICACT20210116 Answer.5
Answer by Auhor shenhanji@cnic.cn 2021-02-10 오후 8:41:46
Chrome Click!! |
| It's an interesting topic. Why is the recall lowered when the NOC is over 50? Doesn't this have anything to do with the accuracy paradox? Although the recall is low, it is questionable whether the high accuracy is a reasonable model. |
This is a very good question. I will pay more attention to the reasons for this result in the next step, and I will do further research |
ICACT20210116 Question.6
Questioner: iglee@sungshin.ac.kr 2021-02-10 오후 8:32:21 |
ICACT20210116 Answer.6
Answer by Auhor shenhanji@cnic.cn 2021-02-10 오후 8:32:21
Chrome Click!! |
| I am curious about the accuracy of detection for untrained websites. In addition, I am wondering if it can detect and learn a new web page tampering method function while learning |
In the future, we can try to add deep learning method to solve this problem.
|
