MODAL-SPLIT MODEL
A. Revealed Preference
Murni
Sintetis : logit
probit
B.Stated Preference
Logit
PEMILIHAN MODA
PENGANTAR
STRUKTUR MODEL
FAKTOR2 PENGARUH
TINGKAT PENYEDERHANAAN
PENGGOLONGAN MODEL
PENJELASAN PER MODEL
TEKNIK STATED PREFERENCE
5.2 Struktur Model
Karakteristik Perjalanan
dan kondisi Sosek
Bangkitan
Pembebanan
5.3 Faktor Yang Mempengaruhi Pemilihan Moda
2. Karakteristik perjalanan
trip length
tujuan perjalanan
waktu saat dilakukannya perjalanan
Karakteristik Umum:
banyak memakai variabel zona atau rumah-tangga, misalnya pemilikan kendaraan, kerapatan pemukiman.
ukuran karakteristiksistem transport biasanya digeneralisasi untuk setiap tujuan, misalnya dengan menggunakan indeks
daya hubung
per-distribusi
dimana :
Aj = daya tarik zona j (dengan angkutan umum atau mobil)
Fij = friksi waktu tempuh dari i ke j = (1/waktu tempuh door-to-door)
n = jumlah zona.
Dengan mengguanakan dua variabel tersebut didapatkan kurva tiga dimensi (berpermukaan), yang menggambarakan
propirsi pemakaian angkutan umum seperti tertera pada Gambar 1.2. Untuk keperluan maksud perjalanan lainnya (belanja
misalnya) dapat dilakukan hal yang sama
Tabel 1. Penyertaan Variabel Dalam Pemilihan Moda
Variable (Attribute) Trip-end model Trip-interchange model
Chicago Pittsburgh Eric Puget Southeastern Washington Twin San Buffalo London
Sound Wisconsin DC Cities Juan
Karakteristik Perjalan:
Jumlah maksud perjalan 2 3 1 4 7 2 3 2 2 3
trip length x
waktu x x
Ke arah pusat kota x
x x
Karakteristik Pengendara:
Pemilikan kendaraan x x x x z x
Kerapatan pemukiman x x x
Pendapatan x x x x
Jumlah pekerja se rumah x
Jalark ke pusat kota x
Kerapatan pekerja x
x1 x2 x3 x4
mobil : 5 0 20 100
bus : 10 15 40 50
Soal 2 :
Dengan situasi sama seperti soal 1, akan dibuat lajur bus terpadu (BT) pada i –j
yang sama.
O/D x1 x2 x3 x4 x1 x2 x3 % mobil
A– U
B–U jalan raya Angkutan
C–U umum
D–U
MODEL BIASA :
% mobil = f ( C )
C = Cmobil - Cumum
Cmobil = x1 + x2 + x3 + x4
Cumum = x1 + x2 + x3
MODEL PERILAKU :
C = a + bx1 + cx2 + dx3 + ex4
Ongkos : - Mobil : C1
- publik : C2 +
exp( C1 )
P1
exp( C1 ) exp (C2 )
1
P1
1 exp (C2 C1 )
exp (C2 C1 )
P2 1 P1
1 exp (C2 C1 )
P1 P 1
1
P2 1 P1 exp (C2 C1
∴ Y =aX+b
REVEALED PREFERENCE MODE
CHOICE LOGIT MODEL
1
2.13 of the “traveler surplus” ? (Hint : Remember that utilities Ūj are themselves defined
up a common constant).
2.14 Show that in case of the illustrative example used in the section 2.3.1, the numerical
values of the probabilities of choice (0.45)–(2.22) may be obtined as the solution of
the R.T.’s U.M. Problem (2.55)-(2.56).
2.15 Show that the avarage received utility Ŵ as specified in Formula (2.22) may be
obtained as the optimal value U * of objective function U in Problems (2.55)-(2.56).
(Hint: Simply replace the ezpression of the optimal demands in the objective function
and simflify).
2.17 Show that the R.T’s U.M. Problem (2.55)-(2.56) always has a unique solution (Hint:
Evaluation the Hessian of objective function and use the fact that all the constraints
are linear).
2.18 Show that when the parameter in logit model (2.11) is equal to 0. the probabilities
are all equal.
2.19 Show that when the parameter in logit model (2.11) is equal to . the probabilities
are all equal to zero,except for one.
This procedure is based on results due to Clark (1956), and will be illustrated in the case of
three alternatives. In this case, one may write Max (V1, V2 V3) = Max { Max (V1, V2), V3 }.
If V1and V2 are both normal variables, which in general may be correlated, then Max (V1, V2)
is approximately normal, with a mean 12 and variance 212. Furthermore, we have
P3 = Pr (V3 > Max (V1, V2)) = Pr (V3 – Max (V1, V2) > 0) (2.62)
in which (V3 > Max (V1, V2)) is the differece between two normally distributed ramdom
variables with known parameters, using the above formulas.
This brings us back to the case of two alternatives (see exercise 2.2). Specifically.
Max (V1, V2, V3, V4) = Max { Max (V1, V2, V3), V4}
so that
P4 = Pr (V4 > Max (V1, V2, V3))
Pj, the probability of alternative j being chosen, is equal to the probability that its random
utility is greater than any of the order alternatives’ random utilities.
It is therefore equal to the probability of the join events A1:
For a given value of ej, since the ej’s are assumed to be independent, the (conditional)
probability of this join event is equal to
Pj = k j P( e k U j e j Uk ; V j
Given the definition of cumulative Gumbel p.d.f. with mean zero and standard deviation,
Formula (A.58a in Appendix A) this is equal to.
Pj = k j P( e k U j e j Uk ; V j
where the positive value of parameter is related to the standard deviation of the ej’s
(Formula (A.59b)). From Formula (A58a) again, the probability density function (x) for ej’
is the derivative of its cumulative probability function:
The unconditional probability Pj of the above event may then be obtained as the expected
value of the conditional probability above, with respect to the distribution of ej:
With a change of variable y = ............, and after some simple manipulations, this is equal
to
with
thus
e Uj
Pj ; Vj
e
k
Uk
P ( V x ) P ( Uj e j x ; V j ) e...teu ka harti...?
j
exp
e x U j E
j
exp e x U j - E
exp exp x - E In
e U j exp - e -.......?
j
with
1
K - E In
e U j
j
According to Formula (A58a), this is the expression of the general cumulative Gumbel
p.d.f. Consequently,using Formula (A59c) the expected value is equal to
E/ K
1
W In e U j
According to Formula (A.69) in Appendix A, given the R.T.’s indirect utility NŴ,where Ŵ is
the individual indirect utility as specified in Formula (2.22) above, the corresponding direct
utility function may be derived as the solution of the minimization problem.
Min W (e j )
(e j )
such that
e Tj
j j T0 Nb
where T0 is the amount spent on other things than travel, (numerraire), and Nb is the total
(R.T.) budget. The problem is then
1
e
( A j - e j )
MinNb N In
(ej )
j
subject to budgetary Constraint (2.67b). The other constraint on the problem is consistency
Constraint16
T j
j N
It is easy to show that problem (2.68), (2.67b), (2.69) is convez, (see section 2 of
Appendix A), and has a unique solution. With the change of variable
1
A j - e j j
Which implies
e e j
1 1
ej A -
In e ; j j
1
T e T
1
j In j A j T0 Nb
j
j
j
Setting its partial derivatives with respect to the ej’s equal to zero, we get
T j
1 1
j
e j
and
1 T j
e j
*
; j
or
* 1 T j
e Aj - In ; j
(2.71)
j
N
- 1
Nb
In
T j
T j In Tj T Aj j T0
j j j
N N -
e In T
1*
Nb In Nb
j
j
j j
(2.72)
The left-hand side of this equality is equal to the optimal value of objective function (2.68),
i,e. the direct utility function. Indeed, replacing e*j by its optimal value above, Formula
(2.71), we have