Anomaly Detection Developer Guide
In essence, the AD is integrated by the following internal components:
- Workflows. Building of Machine learning Models, each workflow perform a machine learning algorithm
- Deployer-Manager. Board to manage the actions for an ML models: Deploy/Stop, Delete, visualization
- Broker. Web Interface of the Broker to explore queues
- Publisher. Machine simulator to publish data on the broker
Controller, An API to perform the machine learning pipeline from datasets importation to Model downloading.
Model Manager. An API to perform the task asked by the Deployer board
Deployer: A REST API to deploy in real-time the models built through the Web-UI
Within the EFPF platform, the AD interacts with the following external components:
- EFPF Portal, the portal provides the secure entry point to the AD interface
- EFPF Data Spine, the Data Spine provides the necessary data brokerage and data transformation support to the AD
The below diagram shows the architecture described above.
The source code is placed in the following link.
To download the source code, create new directory, open a command terminal from the new directory and clone the repository by running the following command:
git clone https://gitlab.fit.fraunhofer.de/efpf-pilots/t42-data-analytics-ice.git