Disruptions data queries

[ciaran haines]

I'm setting up to do some network predictions in the future and am starting to ingest the necessary data. What I'm looking at currently is any predictive power in the free text stuff. My general idea is something using historic delay attribution data and the text from the generating incidents to do something like a delay probability prediction. 

I'm not a noob but I have been ignoring this data for a while. Apologies, but I'm hoping to crowdsources away my inevitable stupid first pass mistakes. 

• The NationalRail Disruptions API seems to be incidents organised per station, and it seems like the wrong way to get the global incidents data. Are the incidents in the NationalRail Disruptions API the same as the incidents in this feed Knowledgebase Incidents data?
• I assume many incidents are closed, but not all?
• Are the incident IDs referenced anywhere else - Darwin timetable updates?
• Is there anything else obvious I'm missing, apart from `free text fields = awful`?

Thank you all in advance for any help!

[Ian Sargent]

Trying to do delay probability prediction is like trying to predict what this week's lottery numbers will be!

The number of factors (many of them oustide of the rail industry's control) that can influence delay are innumerable, so predicting the next delay on what has happened in the past is pointless. 

[ciaran haines]

Hi Ian, Sorry I didn't reply closer to the time - life just got away from me. We're actually predicting passenger crowding. I agree that specific delay prediction is a fool's errand, but predicting the probability of delay has turned out to be fairly informative for improving crowding predictions. Ingesting the  disruptions feed is part of this effort.

[Ian Sargent]

If you want to predict passenger crowding then looking at the dates for major events (sports and otherwise) may give you a better start. Also things like school holidays - as I found out once travelling from Glasgow to Ardrossan on the first week of the summer holidays!

[ciaran haines]

We've already got that covered as best we can. Excellent, thank you!