Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Upcoming SlideShare
×

Better Urban Travel Time Estimates Using Street Network Centrality #Eurocarto15

839 views

Published on

Anita Graser, Maximilian Leodolter and Hannes Koller
Towards Better Urban Travel Time Estimates Using Street Network Centrality

Published in: Technology
• Full Name
Comment goes here.

Are you sure you want to Yes No
Your message goes here
• Be the first to comment

• Be the first to like this

Better Urban Travel Time Estimates Using Street Network Centrality #Eurocarto15

1. 1. Towards Better Urban Travel Time Estimates Using Street Network Centrality Anita Graser, Maximilian Leodolter & Hannes Koller
2. 2. Problem Statement We want to: 1. Predict travel time / speed on a certain road at a certain time 2. Where we don’t have measurements 3. Using only street network data – NO travel demand data 2
3. 3. Travel Time Prediction Usual approaches:  Sensor measurements  Traffic model  Fallback: Speed limit ×correction factor 3
4. 4. 4 Our Basic Prediction Model MAPE Explanatory Variables: - Time of day - Road class - Speed limit
5. 5. Centrality 1. Closeness: “Central”  “Peripheral” How far is it to all other network nodes? 2. Betweenness: “Important”  “Unimportant” How often is this location on the shortest paths between network nodes? 5
6. 6. 6 Global Closeness
7. 7. 7 Local Closeness
8. 8. 8 Global Betweenness
9. 9. 9 Local Betweenness
10. 10. Linear Regression Model Standard Linear OLS (ordinary least square) Regression model Y = X +  (1) where  𝛽 = (𝑋 𝑇 𝑋)−1 𝑋 𝑇 𝑌 (2) Estimators:  𝒚𝒕,𝒇 = 𝜷 𝒕 + 𝜷 𝒇,𝒔 ∙ 𝒔 + 𝜷 𝒇,𝒃 ∙ 𝒃 + 𝜷 𝒇,𝒄 ∙ 𝒄 + 𝜷 𝒇,𝒃𝒄 ∙ 𝒃 ∙ 𝒄 (3) Where  Y … Speed records  X … Model matrix ( daytime t, frc f, betweenness b, closeness c, speed limit s)  𝛽 … stacked vector of OLS estimates 𝛽𝑡, 𝛽𝑓,𝑠, …  b … betweenness  c … closeness  s … speed limit   … Residual 1013.11.2015
11. 11. Centrality Model Parameters  1113.11.2015 Central Peripheral Unimportant Important Secondary & tertiary roads Local roads & residential streets
12. 12. Result1: Improved Coverage of Estimators 1213.11.2015 Secondary & tertiary roads with speed limit 50 kph
13. 13. Interpretation of Results 1. Base model 2. 1st extension using global centrality 3. 2nd extension using local centrality 13
14. 14. 14 Mean Percentage Error Base Model MAPE
15. 15. 15 Mean Percentage Error 2nd Extended Model MAPE
16. 16. MAPE 16 Mean Percentage Error Base Model
17. 17. MAPE 17 Mean Percentage Error 2nd Extended Model
18. 18. 18 Improvements using Global Centrality MAPE Difference 8.5%
19. 19. 19 Improvements using Local Centrality MAPE Difference 14%
20. 20. 20 Difference between Global & Local Centrality MAPE Difference 6%
21. 21. Model Performance Summary 2113.11.2015
22. 22. Contact Anita Graser anita.graser@ait.ac.at @underdarkGIS