Skip to content

A Gradient Boosted Model that triages fraudulent events, and recommends whether or not to flag a given event for further investigation.

Notifications You must be signed in to change notification settings

jeh0753/fraud_detection

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Fraud Detection Case Study

Our team built a Gradient Boosted Model that triages fraudulent events. Our model produces an F1 score (weighted precision and recall) of 0.66.

We identified these factors as key indicators of fraud, and used our available data to find appropriate proxies:

  • Completeness of the Event Listing
  • Country where Event Occurs
  • Veracity of Event Listing
  • Transaction History of Event Organizer
  • Payment Method Requested
  • 'Value' of the Event (Expected Payout Size)
  • Conference vs. Non-Conference

File Structure

Data: The data used in this case study is confidential, and therefore not included in this repository.

Model: Our model is in the home directory, under model.py. Predictions are generated via predict.py

A few notebooks can also be found outlining our feature engineering process, and our model development process.

About

A Gradient Boosted Model that triages fraudulent events, and recommends whether or not to flag a given event for further investigation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published