Skip to content

UKHomeOffice/drt-analytics

Repository files navigation

DRT Analytics App

Introduction

This app produces ML models for arrivals:

  • How many minutes off its scheduled arrival time an arrival will be
  • How many minutes an arrival will take to get to chox from touchdown
  • The likely walk time passengers will have from the plane to the immigration hall

Accuracy

Various models were tested for accuracy against LHR T2 data

6 months of data were used - 5 for training and the most recent 1 month for validation

Off-schedule minutes

The best combination of model partition & features was found to be

Terminal / Origin :: day / am/pm / carrier / flight number

Percentages in the headings refer to improvement in accuracy over the baseline model

Percentages in the table refer numbers of arrivals with the given improvement, eg 35% of arrivals had an improvement of 10% or more

Partition Features 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Term / Origin day of week / am or pm / carrier / flight number 34% 23% 18% 12% 5% 5% 5% - - -
Term / Flight Number / Origin day of week / am or pm 31% 23% 18% 12% 7% 5% 5% - - -
Term / Carrier day of week / am or pm / origin / flight number 23% 17% 11% 8% 2% 2% 2% - - -

Chox minutes

The best combination of model partition & features was found to be

Terminal / Carrier :: day / am/pm / origin

Partition Features 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Term / Carrier day of week / am or pm / origin 88% 73% 47% 29% 17% 8% 5% 2% 2% 2%
Term / Carrier day of week / am or pm / origin / flight number 88% 70% 47% 29% 20% 8% 5% 2% 2% 2%
Term / Flight Number / Origin day of week / am or pm 66% 44% 26% 15% 8% 6% 5% 4% 4% 4%

Walk time minutes

The best combination of model partition & features was found to be

Terminal / Carrier :: day of week / am/pm / origin / flight number

Partition Features 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
Terminal / Carrier day of week / am/pm / origin / flight number 91% 91% 91% 91% 91% 88% 73% 58% 38% 14%
Terminal / Carrier day of week / am/pm / origin 91% 91% 91% 91% 91% 82% 70% 52% 38% 14%
Terminal / Origin day of week + part of day + carrier + flight number 83% 83% 81% 81% 81% 80% 70% 51% 26% 9%

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •