Predict fraud on new events from a stream

info

• Follow the steps mentioned below.
• Total time taken for this task: 10 Minutes.
• Pre-requisites: Fraud detection model is built, and input dataset has data from a stream.

Tidbits

• AWS Fraud Detection service is not natively supported by Amorphic.
• We are going to run an existing Amorphic job which has all the necessary permissions to access AWS Fraud Detector.
• Do not try creating a new Amorphic job. You will run into access issues.

Create an output dataset

• Click on 'DATASETS' --> 'Datasets' from left navigation-bar.
• Click on ➕ icon at the top right corner.
• Enter the following information and click on 'Register'.

{
  "Dataset Name": "fraud_detection_results_<your_userid>"
  "Description": "This dataset is output from fraud detection use-case."
  "Domain": "workshop(workshop)"
  "Data Classifications":
  "Keywords": "S3"
  "Connection Type": "API (default)"
  "File Type": "csv"
  "Target Location": "S3"
  "Update Method": "Append"
  "Enable Malware Detection": "No"
  "Enable AI Services": "No"
  "Enable Data Cleanup": "No"
}

Configure Amorphic Job

• Click on 'ETL' --> 'Jobs' from left navigation-bar.
• You will see a job finace_fraud_detector.
• Click on 'View Details'.
• Click on ✏️ icon at top right corner to edit the job details.
• Click on 'Datasets, Parameters & Libraries' tab.
• Do not remove existing datasets.
• Add fraud_detection_results_<your_userid> to 'Datasets Write Access'.
• Add stream_2_s3_fraud_detection_<your_userid> to 'Datasets Read Access'.
  • Note: If you were not able to add data to stream_2_s3_fraud_detection_<your_userid>, use stream_2_s3_fraud_detection_ankamv dataset.
• Click on 'Update Linked Entities'.

Update Code and run job

• Click on 'Edit Script' icon at top right corner.
• Toggle 'Read mode on' switch.
• Change username. Replace username in input_ds, output_ds, ENTITY_TYPE, EVENT_TYPE, MODEL_NAME, DETECTOR_NAME, DETECTOR_VER variables as shown below.
  • Note: You may download the backup code finace_fraud_detector.py from fraud_detection_resources dataset.

• Click on 'Save & Exit'.
• Click on ▶️ icon at the top right corner to run job. Click on Submit.
• Click on 'Executions' tab.
• Click the 🔄 icon to refresh the status. It finishes in approximately 4 minutes.

Check output

• Once the job is finished, status will turn from 🟠 to ✔️.
• Go to the Amorphic Dataset: fraud_detection_results_<your_userid>. Use the navigator to quickly go to the Dataset page. Press ctrl two times successively and type the dataset name fraud_detection_results_<your_userid>.
• Click on the files tab.
• Click on three dots ⋮ in front of the file to download it.
• Verify the last four columns of the output.

Congratulations!!!

• You have successfully predicted the finance near real-time events.
• You may add the prediction job to the scheduler to run every six minutes for automation.