Retail Fraud Detection

Friday, November 1, 2024

I created and implemented a machine learning model to identify anomalous employee transactions.


Premise

At the start of 2023 a Fortune 500 client sought to modernize a key fraud detection tool. This tool was the centerpiece of the organization’s Loss Prevention department by granting visibility into trends within customer transactions. The client sought to progress the tool by:

  • Implementing a predictive modeling component to the analysis
  • Contextualize trends for each employee and location
  • Segment transactions by cash and credit designation

The success of the project would be measured in terms of recovered dollars and short investigation times. The former criteria is standard for fraud investigation and sometimes required law enforcement intervention. The latter criteria defines the client’s position as a national leader in loss prevention.

Results

  • Reduction in investigation time from 1 month to days
  • 98% return rate of stolen revenue
  • Reduction in total amount returned due to shorter case times

Skills Used

  • Python
  • CI / CD
  • Azure
  • Spark
  • PySpark
  • SQL
  • Machine Learning