Open mmichaelzhang opened 3 years ago
Hi Michael,
Could you tell me which model you use? What steps you have run
Regards, Helen Guo
helen.guo | |
---|---|
@.*** |
Signature is customized by Netease Mail Master
On 10/21/2021 21:24, Michael Zhang wrote:
Hi, I tried to run your latest version code and here is the result I got on HDFS dataset:
best threshold: 0, best threshold ratio: 0.2 TP: 2999, TN: 546622, FP: 6746, FN: 7648 Precision: 30.77%, Recall: 28.17%, F1-measure: 29.41% elapsed_time: 919.4318611621857
It is far away from the reported score in the paper. Can you provide a version that can at least be close to the reported result?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android.
I am running with logbert model.
I first used download_hdfs.sh in main/script/, whish is supposed to download necessary data, but the anomaly_label.csv did not automatically download, so I manually added it.
I then ran init.sh, which creates folders, followed by python data_process.py, python logbert.py vocab, logbert.py train and logbert.py predict.
Also, for the script of BGL dataset, it seems unable to generate parameters.txt, and I have to added it based on the script generated for HDFS.
I also run on HDFS dataset and logbert but got: TP: 8336, TN: 551206, FP: 2162, FN: 2311 Precision: 79.41%, Recall: 78.29%, F1-measure: 78.85% elapsed_time: 1320.057204246521
Can't reach 82% f1 like the original paper
Hi, I tried to run your latest version code and here is the result I got on HDFS dataset:
best threshold: 0, best threshold ratio: 0.2 TP: 2999, TN: 546622, FP: 6746, FN: 7648 Precision: 30.77%, Recall: 28.17%, F1-measure: 29.41% elapsed_time: 919.4318611621857
It is far away from the reported score in the paper. Can you provide a version that can at least be close to the reported result?