16
Mode Choice model USING MULTINOMIAL LOGIT MODEL TRANSPORTATION SYSTEMS ANALYSIS SUMMER 2015

Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Embed Size (px)

Citation preview

Page 1: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Mode Choice modelUSING MULTINOMIAL LOGIT MODEL

TRANSPORTATION SYSTEMS ANALYSISSUMMER 2015

Page 2: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Introduction Mode Choice modelling

◦ Third stage in 4-stage transport modelling

Data : Household travel survey ◦ Variable groups: Socio-economic, Level of Service, Demographic

Location: Windsor, ON◦ High level of vehicle ownership (automotive capital of Canada)◦ Spread out geographically◦ No transit services to suburbs-Lasalle, Amherstberg, Lakeshore etc

Modelling technique: Multinomial Logit model

Software tool: NLOGIT5 (Student version)

Page 3: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Source: www.bikehub.co.uk

ObjectiveFrom the given data, find the variables which have a significant impact on the choice of mode for work-trips and analyse the effect of the variables (positive/negative) on the choice of each mode using a discrete choice method.

Page 4: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Literature Review[Ding et al., 2014 (Exploring the influence of built environment on tour-based commuter mode choice: A cross-classified multilevel modeling approach)]

◦ Distance of home zone from the work location is significant and has a positive effect on auto mode◦ Employment density at work location and population density at home location both significant, but

employment density at work location more so◦ Travel time has a negative impact on auto mode◦ Highly mixed land-use living areas encourage the use of transit for work while mixed land use at

work location not significant[Yong Le Loo et al., 2015 (Transport mode choice in South East Asia: Investigating the relationship between transport users’ perception and travel behaviour in Johor Bahru, Malaysia)]

◦ Variables having a positive effect on public transport use were location of residence, students studying in Singapore, education-trade and technical skills institution and education-post secondary institution

◦ Variables having a negative impact were, gender-female, age(45-54), employed in Johor Bahru and employed in Singapore

Page 5: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Literature Review[Owen A., 2013 (Modeling the commute mode share of transit using continuous accessibility to jobs)]

◦ Transit mode share was found to decrease with increase in household income, increase in population of white, non-hispanics and vehicle ownership.

◦ Household size and education had a negative association with transit ridership.

[de Palma and D Rochat, 2000 (Mode choices for trips to work in Geneva: an empirical analysis)]◦ Variables having a positive impact on number of auto trips: Number of years of commuting,

cross-border travel, duration of daily congestion, weather, female, size of the household, children going to school, young people with age less than 30years

◦ Variables having a negative impact on number of auto trips: Travel time, travel cost, flexible work hours, frequency of congestion, senior people with age more than 50 years, employed in top management, education level

Page 6: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Literature Review [M El-Sayed El-Bany et al., 2014 (Policy sensitive mode choice analysis of Port-Said City, Egypt)]

◦ High income has a positive effect on car/taxi use◦ Out of vehicle travel time has larger impact (negative) than in-vehicle travel time on auto use

[J Zhou, 2012 (Sustainable commute in a car-dominant city: Factors affecting alternative mode choices among university students)]◦ Possessing a discounted transit pass has a positive effect on alternative mode use◦ Commute distance is positively related to carpool. Distance not significant for walking, biking

or transit modes◦ Gender, education level and age significant and positive co-relation to alternate modes

Page 7: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Hypothesis formulation – Data exploration

0 1 2 3 4 50

20

40

60

80

100

120

Number of Vehicles & Mode Share

Walk/BikeTransitAuto

0 1 2 3 4 580

85

90

95

100

105

Number of Bicycles & Mode Share

Walk/BikeTransitAuto

1 2 3 4 5 675

80

85

90

95

100

105

Household size & Mode Share

Walk/BikeTransitAuto

Page 8: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Auto Transit Walk/Bike0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

Employment-type & Mode Share

Full-TimeHome-makerPart-TimeRetiredSelf-EmployedStudentUnemployed

Auto Transit Walk/Bike0.00

10.0020.0030.0040.0050.0060.0070.0080.0090.00

100.00

House-type & Mode Share

ApartmentDuplexSingle-FamilyTownhouseOther

Auto Transit Walk/Bike0.00

20.00

40.00

60.00

80.00

100.00

120.00

Age-group & Mode Share

<=1516-2526-3536-4546-5556-65>65

Hypothesis formulation – Data exploration

Page 9: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Hypothesis formulation – From past research and given data

Household incomeTrip distanceGender-FemaleHousehold sizeVehicles Ownership

Travel CostTravel timeAge 5

Age 6Age 7

Travel CostTravel timeHousehold incomeTrip distanceGender-FemaleVehicle Ownership

Auto Transit

Page 10: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

HypothesisMode

Variable Auto Transit Walk/Bike

Socio-Economi

c

HOUSEHOL +VEHICLES +BICYCLES +GENDER +APT +DUPLEXSING_FAM +THOUSE +OTHERDFULL_TIM +HMAKER +PTIME +RTRDSELFEMPSTUDENT + +UNEMPINC + - -

ModeVariable Auto Transit Walk/Bike

Level of Service

TRP_DISTANCE + - -TT_ATUO -TT_TRANS -TT_WB -Ttime - - -TC - - -

Demographic

AGE1 +AGE2AGE3AGE4AGE5 +AGE6 +AGE7 +

Page 11: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Utility MatrixAlt A1 A2 Vehicles Bicycles Ttime TC Full_tim Student Thouse Sing_fam Age3+Age4+Age5 Vehicles/Househol

AT CA 0 NVEH 0 TT TC FTE 0 TH 0 WORKAGE 0TR 0 CT 0 0 TT TC 0 STDT 0 0 0 NVEHHHTWB 0 0 0 NBIKE TT TC 0 STDW 0 SINGFAM 0 NVEHHHW

Page 12: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Goodness of Fit of modelρ2= 0.34

AT TR WB Total

AT 717 10 28 754

TR 10 1 1 12

WB 27 2 17 46

Total 754 12 46 812

Crosstab: Comparison of actual and model results

Page 13: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Model ResultsProb

95% confidence intervalMODE Coefficient Error z |z|>Z*CA -2.955 0.784 -3.77 0.00 -4.49 -1.42NVEH 1.152 0.330 3.49 0.00 0.51 1.80TT -0.057 0.013 -4.47 0.00 -0.08 -0.03TC -0.349 0.349 -1.00 0.32 -1.03 0.33FTE 0.639 0.479 1.33 0.18 -0.30 1.58TH 1.719 1.156 1.49 0.14 -0.55 3.99WORKAGE 0.689 0.431 1.60 0.11 -0.15 1.53CT -2.656 0.967 -2.75 0.01 -4.55 -0.76NVEHHHT -1.486 1.124 -1.32 0.19 -3.69 0.72STDT 2.138 1.016 2.10 0.04 0.15 4.13NVEHHHW -1.080 0.714 -1.51 0.13 -2.48 0.32NBIKE 0.310 0.133 2.33 0.02 0.05 0.57STDW 1.381 0.847 1.63 0.10 -0.28 3.04SINGFAM -1.574 0.443 -3.55 0.00 -2.44 -0.70

Page 14: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Comparison of results and hypothesis

ModeVariable Auto Transit Walk/Bike

Socio-Economic

HOUSEHOL + VEHICLES + 1.152VEHICLES/HOUSEHOL -1.486 -1.080BICYCLES + 0.310GENDER + APT + DUPLEXSING_FAM + -1.574THOUSE + 1.719OTHERDFULL_TIM + 0.639HMAKER + PTIME + RTRDSELFEMPSTUDENT + 2.138 + 1.381UNEMPINC + - -

Level of Service

TRP_DISTANCE + - - TT_ATUO - TT_TRANS - TT_WB - Ttime - -0.057 - -0.057 - -0.057

TC - -0.349 - -0.349 - -0.349

Demographic

AGE1 + AGE2AGE3AGE4AGE5 +AGE3+AGE4+AGE5 0.689AGE6 + AGE7 +

Page 15: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Simulation Travel times for transit decreased by 50%, and that of auto increased by 25%

Travel cost for transit decreased by 10% and that of auto increased by 10%

Choice

Base Scenario Scenario - Base

% Number % Number % Number

AT 92.86 754 91.0 740 -1.85 -14

TR 1.48 12 3.4 25 1.93 13

WB 5.67 46 5.9 48 0.19 2

Total 100 812 100 813 0.27 1

Page 16: Mode Choice analysis for work trips using Multinomial Logit model for Windsor, Ontario, Canada

Aakash Bagchi (104296114)

Thank You!