• Pattern Recognition in Multi-Source Time Series Data
Integrated video footage, incident data, and traffic flow data to analyze traffic patterns, applying the pruned exact linear time (PELT) algorithm to detect sharp change points in traffic speed and volume over time series.

• Big Data-Driven Machine Learning Modeling
Leveraged crowdsourced data to develop and validate predictive models for real-time pavement condition evaluation. Analyzed key metrics and compared them against a benchmark index. Utilized geographically weighted random forest (GWRF) modeling to quantify and interpret complex relationships within the data, enhancing decision-making accuracy.