/TraceVAE

The source code for "Unsupervised Anomaly Detection on Microservice Traces through Graph VAE" in WWW2023.

Primary LanguagePython

TraceVAE

This is the source code for "Unsupervised Anomaly Detection on Microservice Traces through Graph VAE".

Usage

  1. pip3 install -r requirements.txt.
  2. Convert the dataset with python3 -m tracegnn.cli.data_process preprocess -i [input_path] -o [dataset_path]. The sample dataset is under sample_dataset. (Note: This sample dataset only shows data format and usage, and cannot be used to evaluate model performance. Please replace it with your dataset.) sample:
python3 -m tracegnn.cli.data_process preprocess -i sample_dataset -o sample_dataset
  1. Train the model with bash train.sh [dataset_path]:
bash train.sh sample_dataset
  1. Evaluate the model with bash teset.sh [model_path] [dataset_path]. The default model path is under results/train/models/final.pt:
bash test.sh results/train/models/final.pt sample_dataset