using AI for fraud detection in car sharing
Last December I was visited Pakistan after a two-year break. I took multiple uber trips to my visit my friends and relatives. We preferred payment by credit cash stored in the Uber app.
In a typical Uber trip, the driver keeps his phone GPS for the entire trip duration. After dropping the passenger on their destination, the driver will either search for the next trip or turn their GPS off to stop offering their services.
The anomaly is that the driver reaches the pickup point, cancels the trip, and then turns his GPS off for some hours. Now a genuine reason could be that the passenger was not there, therefore driver had to cancel the trip and call it a day.
However, some drivers would reach the pickup point, negotiate with the passenger to cancel the trip and still take the ride. This way driver was evading the hefty 30% transaction fees uber charges on every trip for using their platform. The driver got away because Uber has no system in place to catch the fraud.
Uber can deploy AI to catch this fraud in real time. If a driver repeats this behavior more than a certain threshold, then Uber can flag him and monitor him closely. In case he is found guilty of abusing the platform, uber can ban him from the platform for some days.
This way Uber can prevent revenue loss and also nudge the drivers to behave more ethically.