Browsing by Author "Zhou, Yun"

Now showing 1 - 3 of 3

Contributing Factors of DUI Recidivism Among First-Time Offenders in North Dakota
(North Dakota State University, 2022) Zhou, Yun
This study explored utilizing tree-based machine learning models to identify associations in a range of 107 factors and DUI recidivism among first-time DUI offenders. Three tree-based machine learning models, Decision Tree, Random Forest, and Gradient Boosting were performed on 12,879 first-time DUI offenders during 2013-2017 using a three-year following period, to classify repeat DUI offenders. Study cohorts include 11,651 drivers without recidivism and 1,228 drivers with recidivism. The models tested 107 variables/predictors, including the driver’s demographic factors, drinking behaviors, traffic violations, crash histories, DUI-related violations, social-economic factors, and health and safety factors based on the driver’s residence. oversampling technique was used to balance two classes in the training data in all three models. The top 15-20 predictors were selected from the feature impact analyses of these predictions. Lastly, multiple logistic regression analyses were performed to quantify the effects of selected factors/predictors on the outcome. Among the three models, Gradient Boosting achieved the best predictions on both the original and oversampled datasets. Oversample techniques did improve prediction performances by roughly 10% on the F1 score for Gradient Boosting. Results coalesced around two findings. First, male drivers with higher BAC values, younger age at first DUI citation, whose first DUI citation took place during the weekday, had at least one low-risk citation within three years before first DUI citation, and lived in counties with lower income inequality ratio and higher violent crime rate were more likely to commit a subsequent DUI offense. Second, male drivers who complied with a BAC test upon arrest, whose first DUI citation took place on a weekday, had at least one low-risk citation within three years before the first DUI citation, lived in a county with a lower income inequality ratio, and higher violent crime rate were more likely to commit a subsequent DUI offense. Findings can be used by stakeholders in implementing and improving DUI prevention strategies. The study is limited to a single state, but the comparison of techniques and their shared findings suggest that a multitude and variety of approaches may be appropriate in future impaired driving prevention research.
Information Asymmetry in Budget Allocation: An Analysis of the Truth-Inducing Incentive Scheme
(North Dakota State University, 2016) Zhou, Yun
Truth-inducing incentive schemes are used to motivate project managers to provide unbiased project information to portfolio manager to reduce information asymmetry between portfolio manager and project managers. To improve the scheme, we identify the proper value of penalty coefficients in the truth-inducing incentive scheme when information asymmetry is present. We first describe the allocation method that achieves budget optimization under certain assumptions and identify the proper coefficients while accounting for the differing perceptions of both portfolio manager and project managers. We report a bound on the ratio between the two penalty coefficients in the truth-inducing incentive scheme and then we conduct a simulation study to narrow down the bound. We conclude that the penalty coefficient for being over budget should be reduced when the portfolio budget is tight and the penalty coefficients should be equivalent to the organizational opportunity costs when the portfolio budget is sufficient.
Review of Usage of Real-World Connected Vehicle Data
(2020) Zhou, Yun; Bridgelall, Raj; Upper Great Plains Transportation Institute
GPS loggers and cameras aboard connected vehicles can produce vast amounts of data. Analysts can mine such data to decipher patterns in vehicle trajectories and driver–vehicle interactions. Ability to process such large-scale data in real time can inform strategies to reduce crashes, improve traffic flow, enhance system operational efficiencies, and reduce environmental impacts. However, connected vehicle technologies are in the very early phases of deployment. Therefore, related datasets are extremely scarce, and the utility of such emerging datasets is largely unknown. This paper provides a comprehensive review of studies that used large-scale connected vehicle data from the United States Department of Transportation Connected Vehicle Safety Pilot Model Deployment program. It is the first and only such dataset available to the public. The data contains real-world information about the operation of connected vehicles that organizations are testing. The paper provides a summary of the available datasets and their organization, and the overall structure and other characteristics of the data captured during pilot deployments. Usage of the data is then classified into three categories: driving pattern identification, development of surrogate safety measures, and improvements in the operation of signalized intersections. Finally, some limitations experienced with the existing datasets are identified.