search

wuyifan18 / DeepLog

Pytorch Implementation of DeepLog.
MIT License
361 stars 154 forks source link

Model Performance is just 88% #49

Open lisn3 opened 3 years ago

lisn3 commented 3 years ago

I trained the model using parameters in the paper for 300 epochs, but the final training accuracy is just 88%, and the testing F1-score is 88%. Could you please give some ideas? Has anyone got a higher performance?

wuyifan18 commented 3 years ago

@lisn3 try to use list to generate the full dataset, just like this: https://github.com/wuyifan18/DeepLog/blob/502aaf05be4c1251b7dc96f6439025c4fc988c66/LogKeyModel_predict.py#L11-L22

lisn3 commented 3 years ago

Thanks, I will try again.

Abraham12580 commented 3 years ago

Thanks, I will try again.

hi，have your performance gotten improved yet?

ldselvera commented 3 years ago

Unfortunately, that will be the highest performance with this implementation. The only way to increase it is by modifying the model and fine tuning parameters.

hayhan commented 2 years ago

Use one-hot vector for input data may help.

gutjuri commented 2 years ago

If you do what wuifan suggested (replacing set with list) you'll get a tremendously increased performance.

elapsed_time: 2877.333s
false positive (FP): 841, false negative (FN): 381, Precision: 95.138%, Recall: 97.737%, F1-measure: 96.420%
Abraham12580 commented 2 years ago

您好，我已经收到你的邮件。祝您工作顺利，天天开心思密达~                                                                 唐子超