457 points by ml_enthusiast 1 year ago flag hide 16 comments
arien1 1 year ago next
Fascinating topic! Predicting flight delays with ML models is both challenging and useful. I wonder what kind of data these models are trained on?
deepblue2 1 year ago next
I believe most models use data such as weather conditions, historical flight records, and airport congestion to predict delays.
arien1 1 year ago next
Great insight, deepblue2! The quality and relevance of data are crucial in developing accurate predictive models.
mlwhiz3 1 year ago next
Absolutely, deepdata7. We actually demonstrated its feasibility with a couple of airlines. It hugely improved operational efficiency and decision-making.
efficientcoder9 1 year ago next
Impressive work by mlwhiz3, deepdata7. Building efficient ML models that utilize real-time data is no small feat!
mlwhiz3 1 year ago next
Much appreciated, efficientcoder9! The progress in ML technology has enabled us to optimize and utilize real-time data effectively. Here’s to more innovations!
mlwhiz3 1 year ago prev next
I recently worked on a similar project and used historical flight data from the Bureau of Transportation Statistics. It turned out to be quite a beneficial dataset.
jetsfan4 1 year ago next
When you mention weather conditions and airport congestion, are you referring to real-time information or historical data, mlwhiz3?
mlwhiz3 1 year ago next
Both in fact! We incorporated real-time weather data from NOAA and historical congestion data from the airports’ operational databases.
deepdata7 1 year ago next
The use of real-time data in such predictive models is intriguing. Have those models been deployed for real-time decision making for airlines, mlwhiz3?
statsguru5 1 year ago prev next
This reminds me of my research on applying Random Forest algorithms for flight delay predictions. It yielded impressive results.
algoenthusiast6 1 year ago next
Random Forest has been used quite extensively in time-series predictions like this. Doesn’t surprise me you got good accuracy, statsguru5.
quickmath8 1 year ago next
How do these models compare to more traditional methods, algoenthusiast6, like regression analysis?
algoenthusiast6 1 year ago next
Great question! While traditional methods get the base job done, the accuracy and precision that ML models offer give them an edge in flight delay predictions, quickmath8.
neural77 1 year ago prev next
Any thoughts on using recurrent neural networks (RNNs) or long-short term memory (LSTM) for this problem, statsguru5?
statsguru5 1 year ago next
RNNs and LSTMs may work for real-time predictions, but one challenge would be obtaining sufficient amounts of recent flight data, neural77.