π This just in: our app predicts your train's delay better than the operator.
-
This just in: our app predicts your train's delay better than the operator. 
Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.Come try it out (still in beta): https://chuuchuu.com/app
-
This just in: our app predicts your train's delay better than the operator. Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.
Come try it out (still in beta): https://chuuchuu.com/app
@chuuchuu@toot.community it needs a gambling option, to bet of whether it is faster or slower than that
-
This just in: our app predicts your train's delay better than the operator. Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.
Come try it out (still in beta): https://chuuchuu.com/app
@chuuchuu it makes sense that you'd be able to do this, considering that train operators usually don't have much reason to put effort into making their estimates accurate...
but also, you also did, like, proper statistics on this, right? (broken clock twice a day and all that) -
@chuuchuu it makes sense that you'd be able to do this, considering that train operators usually don't have much reason to put effort into making their estimates accurate...
but also, you also did, like, proper statistics on this, right? (broken clock twice a day and all that)@mei I'll have to ask our data scientist for the details, but the predictions are based on the delay data we collect. Basically, what's the typical delay you can expect further ahead given the train's current delay.
If you tap the prediction you'll see the % confidence we give to the prediction (we don't show any when we're not confident enough). And we're monitoring this as it goes of course!
-
@mei I'll have to ask our data scientist for the details, but the predictions are based on the delay data we collect. Basically, what's the typical delay you can expect further ahead given the train's current delay.
If you tap the prediction you'll see the % confidence we give to the prediction (we don't show any when we're not confident enough). And we're monitoring this as it goes of course!
@chuuchuu neat! do you do any second-order correlations along the lines of "usually train A would get held up because of train B, but today train B is even more delayed for other reasons so we probably won't accumulate that particular delay"? -
@chuuchuu neat! do you do any second-order correlations along the lines of "usually train A would get held up because of train B, but today train B is even more delayed for other reasons so we probably won't accumulate that particular delay"?
@mei not yet but this is definitely on the roadmap, as it will help us predict delays earlier!
-
This just in: our app predicts your train's delay better than the operator. Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.
Come try it out (still in beta): https://chuuchuu.com/app
@chuuchuu Sounds like a very promising topic for a thesis or other study
-
This just in: our app predicts your train's delay better than the operator. Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.
Come try it out (still in beta): https://chuuchuu.com/app
-
This just in: our app predicts your train's delay better than the operator. Real example from yesterday: we predicted this Berlin-bound train would be 24 minutes late on arrival, while DB still guessed 8 minutes. The train actually arrived 25 minutes late.
Come try it out (still in beta): https://chuuchuu.com/app
@chuuchuu Awesome! Any plans to add ΔeskΓ© drΓ‘hy to the dataset?
From Austria to Upper/East Germany you'll most likely need to change in Prague - and it would be lovely to have that too with live data & prediction
-
@chuuchuu Awesome! Any plans to add ΔeskΓ© drΓ‘hy to the dataset?
From Austria to Upper/East Germany you'll most likely need to change in Prague - and it would be lovely to have that too with live data & prediction
@cyborg42 unfortunately we havenβt found any open delay data for Czechia yet! Please send it our way if you know where to find it
-
R relay@relay.infosec.exchange shared this topic
