My Research
Extracting Mobility Patterns
from Mobile Phone Data
Using anonymous call detail records (CDR) for about 10 million people in Tehran, we have extracted the origin-destination matrices for different time windows. Moreover, by using different machine learning techniques, we have clustered people into different activity groups and identify their home, work, and recreation approximate locations. With such analysis, various accessibility issues have been addressed across the city.
Developing a Hybrid Car-following Model
We have developed a general hybrid framework for integrating data-driven modeling techniques (e.g., ANNs) into traditional mathematical modeling frameworks. With such integration, we are able to improve the models performance and capture nonlinear phenomena, such as reaction time.
Studying Iranian Non-lane-based driving behavior
In this study, which was conducted as my master's thesis, I used image processing in MATLAB to extract vehicles trajectory to study the driving behavior and spatiotemporal interdependencies of traffic flow in Iran as a developing country. By comparing the results with those from developed countries, I found meaningful differences that significantly affect roadways capacity and safety.
Spatial Analysis of Urban Traffic Crashes by Semi-Parametric Geographically Weighted Poisson Regression
Semiparametric geographically weighted Poisson regression (S-GWPR) is employed for the prediction of crash counts in Mashhad, Iran. SGWPR enables capturing the spatial heterogeneity among crash counts data in different regions of the study area. Not surprisingly, our results showed that the proposed model was more successful in predicting crash counts compared to traditional ordinary least squares (OLS) regression models.
Spatiotemporal Analyses of Motorcycle Traffic Accidents Case Study: Mashhad, Iran
This study aimed to detect spatiotemporal autocorrelation in crash count data using the bivariate global (Moran's I) and bivariate local (BLISA) indicators of spatial autocorrelation based on monthly analyses of aggregated motorcycle accidents over 253 traffic analysis zones (TAZ) in Mashhad, Iran. The data were gathered for three successive years (2006-2008), and the results showed that the accidents demonstrated clustered patterns of autocorrelation.
Studying walking duration time and walking mode choice using hazard-based duration and logit models
The study aimed to explore predictors of influencing factors on walking time duration and its choice among individuals in daily trips. For this reason, logit and hazard-based duration models have been applied. The data used in this study have been gathered for updating the comprehensive transportation plan of Mashhad.