Other Research
Research
Transportation data has grown exponentially in the past decades, and opened a new channel to analyze and understand numerous common interests in the domain. At TRUE, we utilize multiple data sources for collision analysis and prediction, including feature selection of injury severity prediction, high collision area classification, and post-collision response.
Risk factor analysis
There were numbers of studies related to risk factor analysis done in TRUE from 2011.
The studies contain :
Risk factor analysis with respect to injury severity
Risk factor analysis with respect to cellphone usage
Risk factor analysis of highway accident with respect to cost severity
Risk factor analysis for fatal crash
Aged pedestrian safety study
Risk factor analysis with respect to injury severity
Many recent road traffic safety studies are focusing on the analysis of risk factors that impact fatality and injury level of traffic accidents. While there are well known factors that affect accident severity such as drug usage and drinking, still there are numerous risk factors that needs to be discovered to affect the severity.
In this research, Naïve Bayes classifier method and the decision tree classifier method were used to reveal the relative importance of the data fields with respect to the resulting severity level.
Some of high-ranking risk factors were found to be having strong interdependency, and it was revealed that only a few numbers of risk factors dominate the severity level.
Risk factor analysis with respect to cellphone usage
Along with the development of IT industry, cellphone usage while driving is becoming one of the major concerns in traffic safety. While cellphone using popularity is growing, there are many cities banning handheld cellphone usage while driving and the efficiency of this banning policy is a controversial topic. By performing turning point analysis technique based on the frequentist and Bayesian approach, it was revealed that the law was critical to the reduction of cellphone-related collisions.
Risk factor analysis of highway accident with respect to cost severity
Current of damage cost calculation of traffic accidents is based only on insurance compensation. This means that the damage cost of road equipment is being ignored, and there is no efficient resolution being prepared to reduce the traffic accidents.
Our research team used a decision tree analysis that considers dependencies between the data fields in order to analyze the relationship between the accident features and the occurrence of damage cost about traffic facilities and road equipment. By using CART algorithm based decision tree classifier, we figured out that cause of accident, accident type, accident location, lane, linearity of road, and road condition are main variables that affects the damage cost. Also, situations that are highly likely to cause high equipment damage cost were extracted from the data.
Other researches related to the risk factor analysis
In addition to the research mentioned above, there are several other risk factor analysis studies such as fatal accident-focused risk analysis, and risk factor analysis about aged pedestrians
Accident hot spot study
There were numbers of studies related to accident hotspot study
DB Fusion
Hotspot clustering study
Factor-wise accident hotspot analysis
DB fusion
Various kinds of traffic-related data are being collected and managed from many agencies. Some of these large-scale traffic DB including traffic volume, flow speed, or public transportation DB are available to public. Accident hotspots (High Collision Concentration Locations, HCCL) researches that only considered accident history DB have limitations in reliability and accuracy due to its insufficient information.
TRUE Lab is working on the fusion of collision history DB from SWITRS with real-time traffic information data from PeMS in order to construct integrated collision DB that provides both environmental factors and driving behavior factors. Combining different kinds of traffic DB will lead to the mutual supplementation of data reliability, offering analysis result of higher credibility.
Factor-wise accident hotspot analysis
Current concept of accident hotspot is explained as HCCL (High Collision Concentration Location), which is determined by counting the accident frequency along absolute post-mile of the road. However, this method does not consider the severity and the cause of the accidents.
In TRUE lab, we focused on factor-wise accident research which enables evaluation of accident in both frequency and severity, using nonparametric models and clustering algorithms.
EMS response time coverage using historic traffic data
In the Emergency Medical Service (EMS), it is widely recognized that the response time - the time from receipt of an emergency call to arrival at the patient location, highly affects the patient survival rate. Response time consists of two components: pre-travel delay and travel time. The pre-travel delay is the amount of time between an emergency call and the vehicle mobilization, while the travel time is the time between the vehicle mobilization and arrival at the patient location.
Between the two components of response time, travel time addresses the larger portion in most cases, and is highly dependent on road traffic condition. There have been numerous researches to analyse the impact of traffic condition to emergency response time based on vehicle speed. However, travel speeds are assumed to be constant in most cases, and variations in travel speed due to conditions common to the road traffic system such as commute hour congestion are not properly addressed.
In this research, time-varying speeds are defined for each street segment within a street network based on the historic traffic data. We propose a GIS-based method to calculate the k-minute travel time contour to represent the response time coverage, incorporating time-of-day and day-of-week effect on travel time in Seoul, South Korea.
Impacts of volcanic eruption on aviation system
The ash cloud from the Eyjafjallajokull volcano in Iceland caused a significant impact on aviation industry on April 14, 2010. Airports across Europe were closed, and at least 17,000 flights a day were cancelled for over a week. Overall 100,000 flights were cancelled and 10 million passengers were unable to board their flights during this week. After Iceland case, there have been many studies trying to address aviation industry impact of Volcanic Ash. However, most of the volcanic eruption scenario studies have limitations.
In TRUE lab, we analyze the damage of the airports and number of the cancelled flights depend on the hypothetical case of Mt. Baegdu eruption model to derive a predicted flights path and diffusion distance of volcanic ash. This research analyzes the effect of volcanic ash to air traffic and airport by quantification value by assessing volcanic ash concentration and its altitude.
