I participated in the Houston ConnectSmart project under Metropia as part of the TxDOT-supported mobility platform team. Each team member focused on a specific aspect of the system; my work was primarily centered on behavioral analysis, data integration, and supporting incentive and communication features design.
Key Contributions
Analyzed commuter behavior patterns using app log and location-based data.
Identified user segments and developed behavioral profiles to support personalized route and message prompts.
Supported feature planning by translating data insights into app design decisions for notifications and incentives.
Collaborated with developers and transportation planners to align technical and user-facing elements.
My contributions supported ongoing improvements to the ConnectSmart platform’s usability and effectiveness in encouraging multimodal travel behavior.
This work sample highlights my contributions to both the Final Report and Technical Memo for the MTC pilot program. I served as the lead analyst and technical writer and was also responsible for coordinating data inputs from multiple departments and synthesizing them into a unified report.
Key Contributions:
Developed the methodology for evaluating mode shift behavior under incentive conditions.
Conducted spatial clustering using DBSCAN to identify user-defined locations from GPS data.
Designed and implemented the randomized controlled trial (RCT) structure and evaluation metrics.
Built and interpreted Multilevel Logistic Regression (MLR) models to analyze treatment effects.
Drafted all analysis sections and compiled inputs from various teams into the full reports.
This effort provided a comprehensive understanding of how targeted behavioral interventions can influence sustainable transportation choices in urban settings.
I led the data analysis and model development for the Incentivizing Active & Shared Transportation Pilot Program, a strategic initiative aimed at reducing greenhouse gas emissions in the Bay Area by promoting sustainable transportation options. Using Multilevel Logistic Regression (MLR) models, I analyzed how personalized incentives influenced travelers’ mode choices, including cycling, walking, and public transit.
The findings showed that cycling adoption increased by 65% for trips between 3 to 10 miles, and walking adoption increased by 22% for trips under 3 miles. The study found that drivers aged 37-56 with access to multiple transportation modes and bicycle ownership were most responsive to the interventions. Key areas such as San Francisco, Contra Costa, and Santa Clara counties showed higher responsiveness to the interventions. Short trips and weekdays were more effective for behavior change, and $3-$5 incentives successfully boosted active transportation and public transit use.
Additionally, I was responsible for preparing the technical report and deliverables for this program, which are available on the MTC website. These documents outline the program’s objectives, methodology, and key results, providing actionable insights for future transportation policies.
Full project documentation:
https://mtc.ca.gov/planning/transportation/regional-transportation-studies/incentivizing-active-shared-transportation-pilot-program
Publish at Transportation Research Part A: Policy and Practice 119, pp. 148-161. (SSCI, SCI) (Impact factor 2017: 3.026)
https://doi.org/10.1016/j.tra.2018.11.005
Public transportation is relatively safe and secure, although less convenient than private modes of transport. However, current trends indicate that, by 2030, road traffic injuries will be the fifth leading cause of death globally. This study proposes an approach for identifying hidden contributors to traffic risk in the major metropolitan cities of Taiwan. Our purpose is to offer a comprehensive econometrical framework, using Hierarchical Linear Modelling (HLM), which highlights important contributors to traffic accident risk at different levels of injuries for public transportation. Four models underlying HLM are used to characterize the traffic accident risk. Our empirical results indicate that random intercept and random slope with interaction of HLM (model 4) is the best model. In addition, there are significant regional differences in traffic accident risk depending on the use of public and private transportation, the length of bus routes, daily average number of bus frequency per route, gender, driving habits, and behaviour. Results show that, when the length of bus routes increases by 50% in a city with well-developed infrastructure, such as Taipei, the accident risk would reduce the crash risk from 1.66 to 1.43 (decreases by 0.23), corresponding to 3450 casualties, and the total accident expense can be reduced by NT$13 billion. If daily average number of bus frequency per route in Taichung increases by 50%, there are almost 3000 fewer casualties, and the accident expense decreases by NT$9.6 billion. The results of this study provide suggestions to the government that developing public transportation can effectively decrease road traffic accident risk and accident expense.
Publish at Accident Analysis & Prevention 119, pp. 16-22. (SSCI, SCI) (Impact factor 2017:2.584)
https://doi.org/10.1016/j.aap.2018.06.011
Traffic violations, particularly drink driving, are a menace to the drivers themselves, and to other road users. Drink driving crashes often cause death or serious injury to the driver. Understanding the recidivism effect factor of drink driving is essential for designing effective countermeasures. This study is based on register-based data from the National Police Agency, Ministry of the Interior of Taiwan and monthly administrative area information from 2012 to 2015 for the entire population. Hence, this study not only focuses on the effect factor and violation differences between recidivists and non-recidivists, but discusses the entire regional characteristics effect for recidivism. The purpose of this study is to offer a comprehensive econometrical framework, using a multilevel random effect logistic model, which highlights important contributors to drink driving recidivism from regional attributes. As the study findings from our empirical results indicate, there are statistically significant differences with drink driving in administrative areas, depending on the number of report on drink driving by police, divorce rate of the population, alcohol consumption, number of community security patrol teams, number of bus trips, and level of education. The results of this study provide suggestions to the government for enhancing community security and developing public transportation, both of which can effectively decrease drink driving recidivism and improve public road safety.