Focus Model Overview

FOCUS MODEL OVERVIEWDenver Regional Council of GovernmentsJune 24, 2011

This presentation Ask Erik, Shahida and me questions

throughout General Concepts Break model into four “stages” Then several steps within each stage Describe each step- it’s inputs and

outputs Review stage inputs and outputs Move onto to next stage

General Concept 1:What is a logit model? 2 minute version

A model that calculates the utility of each choice from a set of discrete choices for a decision maker based on characteristics of the choice and the decision maker Suppose we have a trip we know is 3 miles long,

it’s purpose is to eat a meal, it’s final location is the CBD, the person who made the trip is age 68, the person’s income is $28,000/year. What is the utility for this trip for various travel modes?

We calculate the utility of taking transit for example based on the above information and the cost of transit, the time on transit, etc

General Concept 1: Logit Models

Many types: we use multinomial logit and nested logit

Outcome is a simple closed form probability (not the choice) The choice must be randomly selected using a monte carlo process.

Anything to add? Questions?

General Concept 2:Monte Carlo Simulation

If model only assign probabilities to a choice, how do we get a choice

Monte carlo simulation Suppose you have three choices, one with probability 0.1,

one with probably 0.4 and one with probability 0.5. Arbitarily, you assign choice one the range on the

number line from 0 to 0.1 to choice one, from 0.1 to 0.5 to choice two and 0.5 to 1 to choice three.

Then you generate a pseudo-random number from 0 to 1. If the number lands on the range assigned to the choice:

you pick that choice For example, if you generate the number 0.6784536475,

you would select choice 3

General Concept 3: Tours and Trips

HOME WORK

STORETOUR-BASED MODEL•1 home-based work tour•1 shopping stop

TRIP-BASED MODEL• 1 home-based work trip•1 non-home-based trip•1 home-based non work trip

Outbound, Away From Home Tour Half

Inbound, toward home tour half

General Concept 4:Focus Model Flow: “Four” Steps

STAGE 1: Make

PopulationAnd

Network

STAGE 2:Run GISDK to Mode Choice

STAGE 3: C#

Logit Models to

Create Trips

STAGE 4: GISDK

Assignment

FEEDBACK

General Concept 5:Mechanics: Code Types used in Model

C# Code Logit Model Running;

Model Manager

GISDK CodeSkimming,

Assignment, I-E/E-E DIA Trips

SQL Server DatabaseQueries

Data Storage

Java For Population Synthesizer

General Concept 6: Use of SQL Server

Households, Persons and Points in SQL Server

Persons, Trips, and Tours in SQL Server

General Concept 7:GISDK: OldModel Socio-economic

Inputs

Area Type

Parking Cost

Network Processing & Data Preparation

Highway

Skimming

Transit

Skimming

Trip Generation

Trip Distribution

Mode Choice

Highway

Assignment

Transit Assignment

Highway Network Inputs

Transit NetworkInputs

Time-of-Day

General Concept 8:Why are we doing this anyway?

Data on ANY geography: all data is at a point level Using Most Demographic Characteristics of People Walking, biking, and transit

Vehicle Miles Traveled households in LoDo in 2035 Average Bike Miles per persons age 70+ years old

in 2020 Number of Cars owned by college students

attending CU Boulder in 2015 Average Distance to Work by Restaurant Workers

Review of General Concepts 1. Logit Models are models that make assign probabilities to

a set of choices for an individual from a list of discrete choices.

2. The actual choice is made using a monte carlo process. 3. Travel in the model is made on a tour-level, and then a

trip level. 4. We can divide the model into four stages. 5. We use four types of code in the model: T-SQL, C#,

GISDK, and Java. 6. Much of the input and output data is stored in SQL Server. 7. We still have to run parts of our old GISDK code for path

building, skimming and assignment. 8. We are doing this because we can get much finer detail

and answer planning questions better using the model.

Thinking points before we dive into the steps

How is the new model activity-based? How is it disaggregate?

How does the model actually do all this crazy stuff?

How is the old model different than the new model?

How does the model STILL simplify actual

human behavior?

Focus Model Flow: 28 StepsOutside The Speed Feedback Loop: Run Once-STAGE 1GISDK called from C#:GISDK PreprocessJava:3. Population SynthesizerC#4. PopSyn Output Processor5. Size Sum Variable Calculator

STAGE 2 GISDK called from C#:For DIA, I-E, E-E and Commercial Trips1. DRCOG Multi-Period Highway Preprocess 2. DRCOG Multi-Period Transit Preprocess3. DRCOG Transit Preprocess4. Trip Generation5. Highway and Transit Skimming6. Trip Distribution7. Mode Choice

STAGE 3 C# Regular Trips8. Regular Work Location Choice 19 . Tour Main Mode Choice 9. Regular School Location Choice 20. Tour Time of Day Choice10. Auto Availability 21. Intermediate Stop Generation11. Aggregate Logsum Generation 22. Trip Time of Day Simulation12.Daily Activity Pattern 23. Trip Time Copier13. Exact Number of Tours 24. Intermediate Stop Location14.Work Tour Destination Type 25. Trip Mode Choice15.Work-Based Subtour Generation 26. Trip Time of Day Choice16. Tour Time of Day Simulation 27. Write Trips to TransCAD17. Tour Primary Destination Choice18. Tour Priority Assignment

STAGE 4 GISDK called from C#:28. Highway and Transit Assignment

FEEDBACK

User Interface: How the steps look

Focus Model Flow: Stage 1STAGE 1: Make Populati

onAnd

Network

STAGE 3: C# Logit

Models to

Create Trips

STAGE 4:

GISDKAssignm

ent

FEEDBACK

STAGE 1: Make Population and Network

Java: Population Synthesizer C# to process in database: Size

Sum Variable Calculator; PopSyn Output Processor

GISDK called from C#: GISDK Preprocess

Creating networks for example

Population Synthesizer

ACS or PUMS Disaggregate Data

Aggregate Data that We Need to Match:Economic Forecasts, Land Use Forecasts

Disaggregate Population With the Right Portions Matching the Economic and Land Use Forecasts

Questions?