The University of Melbourne
DEPARTMENT OF INFRASTRUCTURE ENGINEERING
CVEN90063 Transport System Modelling
Subject Project 1: Travel Demand Prediction Models
Due Date: June 6th at 23:59
• This is an individual assignment and accounts for 50% of your final mark.
• Please write a short (half-page) response for each of the following tasks summarizing the
steps you took for each task.
• Please submit any PSQL and Python codes, Excel spreadsheets, data files and most
importantly your model outputs separately for each task.
Task 1: Trip Production Model — 10 pts
Develop a“person”trip production linear regression model based on VISTA data using records
from 2012 to 2016, and only consider weekdays morning-peak trips starting between 5am-9am
from home. Your model must include at least 4 explanatory variables and a constant. Evaluate your
model and discuss the coefficient values and significance.
Task 2: Trip Deterrence Function — 10 pts
Calibrate an exponential trip deterrence function based on trip distances reported in VISTA (2012-
2016) and only including trips made on weekdays morning-peak between 5am-9am from home and
shorter than 50kms. (Hint: you can use the field‘cumdist’in table t, and round it to the closest
kilometre).
Task 3: Trip Distribution Model — 20 pts
Calculate the morning peak travel O-D matrix for weekday at the SA2 zone level (309*309) using
the gravity model implementation in python. Use the trip deterrence function from task 2 and
assume that every individual produces 1 home-based trip in the morning and every job attracts 1
home-based trip in the morning peak. Submit your final O-D matrix in a CSV file and report your
total error in matching the zone productions and attractions.
Task 4: Model Choice Model — 30 pts
Develop a Multinomial Logit Mode Choice model based on VISTA trips recorded between 2012-
- Your utility function attributes must all be significant, and you should provide a reasonable
interpretation (for the sign) of the estimated coefficients in your model.
Your model’s goodness of fit measure must be greater than 0.700 and it must include at least one
attribute variable from every category listed below: - Alternative specific constant
- Travel time attribute (for all modes)
- Person attribute (decision maker) scaler variable
- Person attribute (decision maker) categorical variable
- Household attribute of the decision maker (either scaler or categorical)
Task 5: Nested Logit Model — 30 pts
Take one of the best Multinomial Logit Mode Choice model that you developed in Task 4 as the
baseline model. Keep the utility function same. Develop a Nested Logit Mode Choice model.
Evaluate the model coefficients and nest parameters.
Your model’s goodness of fit measure must be improved comparing to the baseline model. Briefly
discuss why you believe that the nested model can make a better prediction.
Plagiarism declaration
By submitting work for assessment I hereby declare that I understand the University’s policy
on academic integrity and that the work submitted is original and solely my work, and that I have
not been assisted by any other person (collusion) apart from where the submitted work is for a
designated collaborative task, in which case the individual contributions are indicated. I also
declare that I have not used any sources without proper acknowledgment (plagiarism). Where the
submitted work is a computer program or code, I further declare that any copied code is declared in
comments identifying the source at the start of the program or in a header file, that comments
inline identify the start and end of the copied code, and that any modifications to code sources
elsewhere are commented upon as to the nature of the modification.