This project translates the Hazelcast Fraud Detection Example from https://github.com/hazelcast/fraud-detection-onnx to Apama and Python. Example data and model are taken from the Github project.
The project reads transaction data from a CSV and publishes them as JSON on MQTT Apama will read in the transaction data from the MQTT broker. The context data (customers, merchants) are read directly from CSV files.
The project runs an inference on each transaction and will output the CreditCard number for each potential fraud.
Before running this project you must install ONNX Engine
- Open an Apama Command Prompt
- run : pip install numpy onnxruntime
install a local broker (e.g. mosquitto)
- Install Jmeter
- Install Jmeter MQTT plugin:
- download jar from https://github.com/emqx/mqtt-jmeter/releases
- put jar into apache-jmeter-\lib\ext
- Load jmx file
- Make sure broker port matches your installation
- Adjust the path in the "CSV Data Set Config" step to the transaction file
- Run Apama project
- Run JMeter script