Fraud Detection_​IQR_​EmailAuto(ToShare)

Sales Transaction Anomaly Detection - Initial Screening & Automated Email Alert

• The Table Creator node is intentionally used for sharing purposes. In real applications, other reader nodes can be used instead.

• In the Numeric Outliers node, use:

  • R7 for QUARTILE.INC

  • R6 for QUARTILE.EXC

• In the Numeric Outliers node, handling multiple subgroups is critical for accurate outlier detection. Detecting outliers on the overall dataset may not identify the true anomalies.

  For example:

  • Sales values for Mouse products may naturally be much lower.

  • Sales values for Laptop products may naturally be much higher.

  If both are analyzed together:

  • Mouse transactions may incorrectly be flagged as low outliers.

  • Laptop transactions may incorrectly be flagged as high outliers.

  This leads to misleading results rather than identifying real outliers. Stratifying data into subgroups is the key approach. For heterogeneous datasets, the Group setting should be used.

• The Table Writer node is intentionally used for sharing purposes. In real applications, other reader nodes can be used instead.

• Copy the "path" from the Table Writer node:

  • Right-click the Table Writer node

  • Select Configure flow variables

  • Paste the copied value into the "path" variable box

• The Email Connector (labs) node requires:

  • Email server address

  • Port number

  • Email address

  • Login credentials

  • Other related email server configurations

• The Email Sender (labs) node requires the necessary information to send emails. The Attachment (input column) should be connected with the path from the Table Reader node.

• Due to their nature, the configuration settings of the Email Connector and Email Sender nodes cannot be shared directly. However, these workflows have been successfully tested and implemented in real-world scenarios using this approach.