Combining Visual Analytics and Machine Learning for Route Choice - - PowerPoint PPT Presentation

▶
combining visual analytics and machine learning for route
SMART_READER_LITE
LIVE PREVIEW

Combining Visual Analytics and Machine Learning for Route Choice - - PowerPoint PPT Presentation

Aug 23, 2023 222 likes •462 views

Combining Visual Analytics and Machine Learning for Route Choice Prediction Application to Pre Tactical Traffic Forecast Rodrigo Marcos Data Scientist, Nommon Solutions and Technologies SESAR Innovation Days Beograd, 29 th November 2017 Scope

slide-1
SLIDE 1

Rodrigo Marcos Data Scientist, Nommon Solutions and Technologies

Combining Visual Analytics and Machine Learning for Route Choice Prediction

Application to Pre‐Tactical Traffic Forecast

SESAR Innovation Days Beograd, 29th November 2017

slide-2
SLIDE 2

Scope and Objectives

Problem:

  • ATFCM in the pre‐tactical phase

Current approach:

  • Based on similarity

http://www.eurocontrol.int/articles/ddr‐pre‐tactical‐traffic‐forecast

Objectives:

  • Use visual analytics to extract route choice determinants
  • Model behaviour of airlines regarding route choice between airport pairs

using machine learning techniques

  • Evaluate pre‐tactical prediction power

2 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-3
SLIDE 3

State of the Art Airline Route Choice Behaviour

Abundant research on tactical trajectory prediction:

  • Prediction of arrival time
  • Conflict detection
  • …

Limited research on airline route choice prediction before the availability of flight plans (pre‐tactical forecast):

  • Luis Delgado (2015) “European route choice determinants”

3 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-4
SLIDE 4

Approach

  • Data: actual trajectories (M3) from DDR2
  • Route clustering per OD
  • Visual exploration of route choice determinants
  • Train a machine learning model
  • Evaluate quality of predictions vs null model

4 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-5
SLIDE 5

Case Studies

  • ODs:
  • Istanbul to Paris
  • Canary Islands to London
  • Multinomial regression
  • Candidate variables
  • Route length
  • Charges
  • Time
  • Schedule
  • Congestion
  • Temporal scope:
  • Training/exploration: AIRACs 1601‐1603
  • Testing: AIRACs 1501, 1502

5 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-6
SLIDE 6

Clustering

6 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Cluster No of flights 139 1 110 2 190 3 218 4 117 5 73 6 29 7 24

Clustered with DBScan Metric: Flown kilometres per ANSP

slide-7
SLIDE 7

Visual Exploration Cost‐worthiness

2 variables considered

  • Average route length
  • Average route charges

1 variable discarded

  • Average flight time

7 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Charges Length Time

slide-8
SLIDE 8

Visual Exploration Airline Behaviour

2 variables considered

  • Arrival time
  • Airline

8 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

22:00‐00:00 (all airlines) 20:00‐22:00 (all airlines) OHY THY

slide-9
SLIDE 9

Visual Exploration Congestion

1 variable considered

  • Average number of regulated

flights 1 variable discarded

  • Average standard deviation of

en‐route FL with respect to RFL

9 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Regulations FL deviation 12:00‐16:00 16:00‐20:00

slide-10
SLIDE 10

Visual Exploration Cluster Properties

10 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Cluster

No of flights

Average length (NM) Average charges (EUR) Regulations per flight

139

1277 1188 0.15

1

110

1314 1144 0.11

2

190

1273 1199 0.06

3

218

1274 1203 0.06

4

117

1256 1207 0.07

5

73

1274 1204 0.1

6

29

1271 1229 0.03

7

24

1304 1152 0.04

Istanbul ‐ Paris

slide-11
SLIDE 11

Visual Exploration Cluster Properties

11 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Canary Islands ‐ London

Cluster

No of flights

Average length (NM) Average charges (EUR) Regulations per flight 659 1620 1653 0.18 1 238 1638 1676 0.13 2 68 1740 1051 0.13 3 13 1732 1582 0.46 4 7 1724 1893 0.42 5 10 1780 1165

slide-12
SLIDE 12

Approach Parameters

Route parameters (used for modelling):

  • Cost‐worthiness:
  • Average route charges
  • Average route length
  • Congestion:
  • Rate of regulated flights

Flight parameters (used for segmentation):

  • Airline (CASK)
  • Arrival time

12 SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-13
SLIDE 13
  • Cost‐worthiness:
  • Average route charges
  • Average route length
  • Congestion:
  • Rate of regulated flights

Modelling Approach Multinomial Regression Model

13

Model of class i and cluster j

  • Xj vector of parameters of cluster j
  • βi vector of constants of model i
  • αj independent constant of cluster j

Variables:

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-14
SLIDE 14

Approach Training and Validation

14

Data Training 70%

Validation

30% 4 3 2 1 Model 4 Model 3 Model 3 Model 2 Model 1 Model 0 2 Guess 2 Guess 0 Guess 0 Segmentation Train model Model validation Routes Clustering ¿=?

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-15
SLIDE 15

Validation Results Canary Islands‐London

15

  • Low number of routes
  • Very different
  • Well explained

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-16
SLIDE 16

Validation Results Istanbul‐Paris

16

  • High number of options
  • Similar routes
  • Missing explanatory variables?

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

Cluster

No of flights

Average length (NM) Average charges (EUR) Regulations per flight

3

218

1274 1203 0.06

4

117

1256 1207 0.07

Cluster

No of flights

Average length (NM) Average charges (EUR) Regulations per flight

139

1277 1188 0.15

5

73

1274 1204 0.11

slide-17
SLIDE 17

Approach Testing

17

Dataset 2 Model 4 Model 3 Model 3 Model 2 Model 1 Model 0 Routes Clustering Dataset 1 Class 4 Class 3 Class 2 Class 1 Class 0 Segmentation Train Predict Route 2 Route 1 Route 0 Compare

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-18
SLIDE 18

Testing Results Canary Islands‐London

18

  • The model captures:
  • behaviour of new airline (Norwegian)
  • airlines changing route options
  • Improvements with respect to null model

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-19
SLIDE 19

Testing Results Istanbul‐Paris

19

  • The model captures:
  • ther routes considered (7)
  • significant change in charges
  • Much better than null model

Cluster Charges (train) Charges (testing) Regulations (train) Regulations (testing) 1188 1305 0.15 0.04 3 1204 1260 0.07 0.02

12:00‐16:00

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-20
SLIDE 20
  • Potential for pre‐tactical demand forecast
  • Range of applicability needs to be clearly identified:
  • Training data requirements
  • Prediction error measurement
  • Generalisation to other ODs

20

Applicability

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-21
SLIDE 21
  • Better explanatory variables
  • Other indicators
  • Congestion as a function of time
  • Other flight inputs: wind, type of regulation, route

availability…

  • Training with several years’ data
  • Continuous training/prediction (automatic adaptive training data)
  • Combination with model‐based approaches (cost optimisation)

21

Future Research Directions

SIDs, Beograd, 29th November 2017 – Combining Visual Analytics and Machine Learning for Route Choice Prediction

slide-22
SLIDE 22

This project has received funding from the SESAR Joint Undertaking under the European Union’s Horizon 2020 research and innovation programme under grant agreement No 699303

The opinions expressed herein reflect the author’s view only. Under no circumstances shall the SESAR Joint Undertaking be responsible for any use that may be made of the information contained herein.

Thank you very much for your attention!

SIDs, Beograd, 29th November 2017 Combining Visual Analytics and Machine Learning for Route Choice Prediction Application to Pre‐Tactical Traffic Forecast