Development of population synthesis and activity pattern generator for activity based travel demand model in Thailand

Name: Natachai Wongchavalidkul

keyword: Activity based travel demand model

LCSH: Transportation engineering -- Computer simulation

; Population synthesis

; Activity generator

Abstract: This research aims at developing the methodology and framework of the activity based travel demand model, specifically for Thailand’s transportation and social characteristics. In the context of activity based travel demand modeling, the research mainly concentrates on the problems of “estimating baseline population distribution” and “simulating individual activities”. To estimate baseline population distribution, an incomplete marginal distribution of population data is considered as the main problem for population synthesis in Thailand. To solve the problem, the optimization of least square errors using Non-Linear Programming (NLP) was proposed in the research. Several proposed objective functions were evaluated using data from 2005 Phitsanulok province’s population demographic survey conducted by Office of Transport and Traffic Policy and Planning (OTP). As the results, the least squares optimization procedure with the objective function on conditional distribution was found to be appropriately applied to estimate the baseline population distribution in the area where marginal distributions of population data are not completely available. Additionally, for the activity generation, the research focused on the improvement of the Agent Based Modeling systems (ABM) which is expected to be more applicable on the future development of multi-agent system. The applications of artificial intelligence and data mining methods were explored in the research. Three proposed modeling approaches, including a rule base approach, a probabilistic approach, and a decision making approach, were evaluated. •In the rule base approach, the research concentrated on the improvement of Classification And Regression Tree (CART) model using in Transportation Analysis SIMulation System (TRANSIMS). Instead of creating the classification tree based on household and householder characteristics. The proposed CART model was developed to classify the individual person based on their socio-economic characteristics. Both continuous and discrete variables were considered in the proposed model. Further, for the splitting criteria, the research proposed the integration of Sequence Alignment Methods (SAM) in CART model which help to improve the classification ability in considerations of activity sequential patterns. •In the probabilistic approach, instead of randomly select activity patterns from the classified database, results from CART, the Bayesian network is applied to generate both activity and its time durations as the probabilistic aspects given the group of classified population, additional individual characteristics, and activity schedule characteristics. From the results, it was found that the model could generate various activity patterns and reduce the limitation of activity pattern selections in the CART model. •In the decision making approach, the Influence Diagram was proposed to extend the Bayesian network model (the probabilistic modeling approach) to the decision approach which is able to consider both probabilities and utilities of activities over time periods. The Particle Swarm Optimization (PSO) was applied to calibrate the utility values used in the model. The individual activity decisions were selected based on the expected utility distribution given by the model. These three modeling approaches were evaluated using the time use survey data for Bangkok’s population from National Statistical Office (NSO). From the results, it found that to generate the pre-planned activity schedule, the model of CART, with the integration of SAM, gave the best match between the generated sequences and the sequences from the survey data. However, the limitation on the choice sets which occur to be problems on the activity reschedule process remains. In this case, the influence diagram model (the decision making approach) gave a flexible framework to reschedule activities over the time periods. The model also provided a better decision behavior framework than the model from rule base approach and the probabilistic approach. Finally, using the existing available data, results on both “estimating baseline population distribution” and “simulating individual activities” creating in this research revealed good modeling outcomes. Hence, the proposed model is recommended to the activity based travel demand model development in Thailand

Thammasat University. Thammasat University Library

Address: BANGKOK

Email: preserv@tu.ac.th

Name: Mongkut Piantanakulchai

Role: advisor

Name: Pruettha Nanakorn

Role: co-advisor

Created: 2015

Modified: 2021-10-28

Issued: 2021-10-28

วิทยานิพนธ์/Thesis

application/pdf

DOI: 10.14457/TU.the.2015.1527

eng

DegreeName: Doctor of Philosophy

Level: Doctoral Degree

Descipline: Engineering

Grantor: Thammasat University

©copyrights Thammasat University

ลำดับที่.	ชื่อแฟ้มข้อมูล	ขนาดแฟ้มข้อมูล	จำนวนเข้าถึง	วัน-เวลาเข้าถึงล่าสุด
1	4800natachai.pdf	2.65 MB	1	2022-02-21 21:16:48

ใช้เวลา

0.036137 วินาที