Transit Demand Forecast

1
Transit Demand Forecast

• Feb 12, 2007

2
Outline

• Four-step modeling
• Discrete choice models
• Assignment 1 (due 2/21)

3
Transportation System
Source Michael McNally, The Four-step Model, 2000
4
Four-step Demand Procedure
SF Bay Area
5
Trip Generation

• Estimate the number of trips originating from or
  destined to each zone (e.g. TAZ) of a studied
  area by relating frequency of trips to
  characteristics of individuals, zone, transport
  network
• Primary tools
• Cross-classification homogeneous groups with
  trip rate available look-up tables
• Regression Models aggregate disaggregate
• Discrete Choice Models
• Trip balancing required

6
Trip Distribution

• Predict spatial pattern of trips between O D
• Primary Tools
• Growth factor (Fratar) model
• Gravity models
• Intervening opportunity model

7
Modal Split

• Predict the shares or number of trips by mode
  (transit vs. auto vs. others)
• Primary tools
• Trip interchange (diversion curve,
  cross-classification, regression)
• Discrete choice

8
Route Assignment

• Assign trips to specific routes on network
• Primary tools Shortest path (time, distance,
  cost)
• All-or-nothing
• Assignment curves
• User Equilibrium with link performance

9
Weaknesses of the 4-step Process

• Trend extrapolation
• Fixed-demand independent of transportation system
• Data requirements
• Aggregation error
• Trip cost components
• Definition of trips

10
Discrete Choice Logit Model
11
Logit Model Example
12
Aggregation Bias

• See spreadsheet

13
Independence from Irrelevant Alternatives

• Logit model assumes that the error terms of the
  utility functions are independent and identically
  Gumbel distributed
• Red-bus/Blue-bus paradox

14
Nested Logit Model

• Partition the choice set into nests
• Utility function for an alternative in a nest
• Composite utility function for a nest
• Probability model

15
SF Bay AreaMode Choice for Home-based Work