Neural Network Modelling of Constrained Spatial Interaction Flows



992 samples into three subsets12:    the    training    [in-sample]    set

M1 = Kxu 1,yu 1 j withu1 =1,...,U1 = 496patterns}, the internal validation set
M2 = {(xu2,yu2 ) with u2 = 1,...,U2 = 248 patterns} and the testing [prediction, out-of-
sample] set
M3 = (xu3,yu3) with u3 = 1,...,U3 = 248 patterns}. M1 is used only for
parameter estimation, while
M2 for validation. The generalisation performance of the
model is assessed on the testing set
M3 .

Though the simplicity of this method is appealing, an obvious concern is the necessary
reduction in the amount of training data. In deciding how to partition the data, a
compromise has been made between creating a test set large enough to fully test the
fitted model while still retaining a sufficient amount of training and internal validation
data. If the test set is too small then the variance of the prediction error estimate will be
high due to the small sample size. Though random splits are commonly used and appear
to work reasonably well in the case of unconstrained spatial interaction, a more
systematic splitting method had to be used in the case of constrained spatial interaction.

Table 1: Descriptive Statistics: The Training, Validation and Testing Sets

Variables

Mean

Standard

Deviation

Minimum

Maximum

Whole Set M

sj

26,364,563

50,350,660

2,310,400

285,193,984

dij

229.4

124.6

30.0

630.0

t
tij

8.6

22.6

0.0

257.9

ti .________________________________

266.0

350.1

41.9

1830.1

Training Set M1

sj

26,142,923

49,711,907

2,310,400

285,193,984

dij

234.1

129.6

35.0

630.0

tij

9.6

26.2

0.0

257.9

ti.________________________________

297.0

429.1

41.9

1830.1

Validation Set M2

Sj

26,517,946

50,891,071

2,310,400

285,193,984

d
ij

219.3

121.4

30.0

590.0

t

tij

7.1

16.6

0.0

166.8

ti.

220.9

221.4

45.6

759.8

Testing Set M3

Sj

26,654,459

51,069,577

2,310,400

285,193,984

Dij

230.3

116.7

37.0

627.0

t
tij

8.0

19.7

0.0

195.2

ti.

249.0

262.5

55.3

895.7

Note: M consists of 992 patterns, M1 of 496 patterns, M2 of 248 patterns and M3 of 248
patterns.

23



More intriguing information

1. Population ageing, taxation, pensions and health costs, CHERE Working Paper 2007/10
2. Fortschritte bei der Exportorientierung von Dienstleistungsunternehmen
3. The name is absent
4. The name is absent
5. Should Local Public Employment Services be Merged with the Local Social Benefit Administrations?
6. Aktive Klienten - Aktive Politik? (Wie) Läßt sich dauerhafte Unabhängigkeit von Sozialhilfe erreichen? Ein Literaturbericht
7. The magnitude and Cyclical Behavior of Financial Market Frictions
8. Institutions, Social Norms, and Bargaining Power: An Analysis of Individual Leisure Time in Couple Households
9. Modelling Transport in an Interregional General Equilibrium Model with Externalities
10. The name is absent
11. Transfer from primary school to secondary school
12. The name is absent
13. Prevalence of exclusive breastfeeding and its determinants in first 6 months of life: A prospective study
14. Manufacturing Earnings and Cycles: New Evidence
15. Who runs the IFIs?
16. CONSUMER PERCEPTION ON ALTERNATIVE POULTRY
17. Uncertain Productivity Growth and the Choice between FDI and Export
18. Howard Gardner : the myth of Multiple Intelligences
19. New issues in Indian macro policy.
20. Does adult education at upper secondary level influence annual wage earnings?