Research

Research Interests:

  • Online experimentation, large scale A/B test
  • Statistical Risk Modeling
  • Statistical Methods in Transportation/Roadway Safety
  • Modeling and Prediction of Rare Events
  • Control Charts for Rare Event Monitoring
  • Online Payment Fraud Detection

Publication:

  • Mao, H., Deng, X., Lord, D., Guo, F. (2019). Adjusting Finite Sample Bias for Poisson and Negative Binomial Regression in Traffic Safety Modeling. Accident Analysis & Prevention, 131, 112-121.
  • Shen, S., Mao, H., Deng, X. (2019). An EM-Algorithm Approach to Open Challenges on Correlation of Intermediate and Final Measurements. Quality Engineering, 31(3), 505-510.
  • Shen, S., Mao, H., Deng, X. (2019). Rejoinder two, QE V 31, I3. Quality Engineering, 31(3)
  • Mao, H., Liu, Y. W., Jia, Y., & Nanduri, J. (2018). Adaptive Fraud Detection System Using Dynamic Risk Features. Preprint: arXiv:1810.04654.
  • Mao, H., Jia, Y., Wang, J., Nanduri, J. (2017). Handling Concept Drift with Self-adaptive Fraud Detection System. Microsoft Journal of Applied Research, 8, 163-172. [Featured on The Journal Cover].
  • Mao, H., Ginther, C. N., Woodall, W. H. (2017). An Evaluation of Wheeler's Method for Monitoring the Rate of Rare Events. Quality and Reliability Engineering International, 33(3), 503-513.
  • Hickman, J. S., Mabry, E., Marburg, L., Guo, F., Mao, H., Hanowski, R., Whiteman, J., Herbert, W. Commercial Driver Safety Risk Factors (CDSRF). No. FMCSA-RRR-17-014. United States. Department of Transportation. Federal Motor Carrier Safety Administration, 2020. (link)

Patent:

  • Jia, Y., Mao, H., Wang, S. J., Marcjan, C., Nanduri, J. N. Hierarchical Profiling Inputs and Self-adaptive Fraud Detection System. (2020). U.S. Patent No. 10,552,837. Washington, DC: U.S. Patent and Trademark Office. (link)

Conference Presentations:

Projects:

Ride-sharing drivers' risk prediction system with driving behavior profile, Didi

Major researcher

This project established an accurate, robust, and explainable driver risk prediction system based on drivers’ driving behavior profile. This is still an on-going project. I cannot disclose more details.

Big Data Methodologies for Simplifying Traffic Safety Analyses, Safe-D

Research Assistant

This study utilizes the kinematic information provided by SHRP2 Naturalistic Driving Study (NDS) to evaluate crash risk on driver-level and trip-level. The SHRP2 NDS is the largest NDS to date and provides a unique opportunity to answer pressing research questions. The SHRP2 NDS included 3,400 primary drivers and collected more than 35 million miles of continuous driving data. More than 1,500 crashes and thousands of near-crashes have been identified in the SHRP2 NDS data.

Handling Concept Drift with Self-adaptive Fraud Detection System, Microsoft

Data Scientist Intern

One major challenge facing payment credit card fraud detection system (FDS) is concept drift. To handle this problem, we proposed using entity profile with fraud feedback to create dynamic risk features, and fundamentally explained why the model with new features is more robust to concept drift by putting everything into statistical learning framework. Real data experiments also show an improvement in ROC curve and Profit Efficiency (PE) measure.

Commercial Driver Individual Differences Study (CDIDS), FMCSA

Data Technician and Statistician

  • Build a relational database for a total of 20,000+ truck drivers, including their three-year driving records, medical examinations, and demographic information
  • Identify significant risk factors and estimate crash risk for each individual by using Poisson regression and survival analysis.

Hours of Service(HOS) Rules Impact Analysis, FMCSA

Data Technician and Statistician

Evaluated the impact of the 2-night restart provision in safety, driver schedules, highway usage, industry costs and productivity, and other aspects.

Home

  • Latest news

© Huiying Mao 2018 | Twitter | Linkedin