MSTDF-AD: Modeling Spatiotemporal Dependency Fusion for Nonstationary Time Series Anomaly Detection

Get Started

1. Install Python 3.11, PyTorch 2.2.1.

pip install -r requirements.txt

2. Download data.
3. Train and evaluate. You can reproduce the experiment results as follows:

bash ./script/run.sh

Download data set

PSM

The dataset can be downloaded at:

https://github.com/eBay/RANSynCoders/tree/main/data

SWaT

SWaT datasets can be obtained by filling out the following form:

https://docs.google.com/forms/d/1GOLYXa7TX0KlayqugUOOPMvbcwSQiGNMOjHuNqKcieA/viewform?edit_requested=true

SMD

The dataset can be downloaded at:

https://github.com/NetManAIOps/OmniAnomaly/tree/master/ServerMachineDataset

MSL and SMAP

The dataset can be downloaded in the following ways

labeled_anomalies.csv: Data processing and data separation between the two spacecraft depend on this file

wget https://s3-us-west-2.amazonaws.com/telemanom/data.zip
wget https://raw.githubusercontent.com/khundman/telemanom/master/labeled_anomalies.csv

Or the dataset can be downloaded at:

https://pds-atmospheres.nmsu.edu/data_and_services/atmospheres_data/Mars/Mars.html

https://nsidc.org/data/smap/data

Download the data set and put it in the corresponding folder, then run

python make_pk.py