Code for the paper MetaAD: A Prototype-oriented Meta Anomaly Detection Framework for Multivariate Time Series

Link (ICML2023)

To run with local environment

Install dependencies using python 3.8+ :

pip install -r requirements.txt 

Run the model:

python runner10.py --test_train_step2 10

test_train_step2 is the number of data used for meta-training.

Data

SMD：

Download the processed data form Google cloud to './SMD/'. SMD Data。

MSL：

Run the data process code in the './MSL/'.

python ./MSL/data_press_1.py