Table 5 Predicting the urban facility accessibility of the neighborhoods in Chicago MSA by transferring the knowledge from other MSAs.

From: Counterfactual mobility network embedding reveals prevalent accessibility gaps in U.S. cities

Category

Input

NY

LA

Dal

Hou

DC

Art & Recreation

mobility statistics

8.31%

6.36%

2.89%

7.80%

4.92%

 

LINE embedding

8.04%

11.79%

14.65%

11.12%

14.11%

 

node2vec embedding

18.05%

26.98%

26.38%

21.50%

21.72%

 

CRANE embedding

30.06%

30.26%

30.34%

26.52%

31.13%

Sports

mobility statistics

5.46%

5.28%

4.75%

5.33%

5.14%

 

LINE embedding

6.81%

6.75%

6.92%

6.66%

6.73%

 

node2vec embedding

6.93%

6.85%

7.17%

7.48%

6.86%

 

CRANE embedding

7.80%

7.89%

7.71%

7.54%

7.61%

Education

mobility statistics

12.98%

13.75%

11.97%

12.81%

14.16%

 

LINE embedding

29.31%

25.51%

31.35%

29.31%

33.07%

 

node2vec embedding

28.41%

27.40%

31.65

29.30%

30.84%

 

CRANE embedding

29.59%

30.00%

30.81%

29.41%

27.12%

Health

mobility statistics

10.90%

15.70%

12.42%

17.45%

15.09%

 

LINE embedding

21.47%

20.46%

19.83%

21.16%

15.37%

 

node2vec embedding

21.65%

18.62%

18.30%

23.49%

13.86%

 

CRANE embedding

22.13%

22.04%

14.42%

11.43%

15.41%

  1. The values show the relative improvement in explained variance compared to using raw demographic features as model input. Bold text indicates the most improved method for prediction performance.