Subasish Das

Investigation on the wrong way driving crash patterns using multiple correspondence analysis

Wrong way driving (WWD) has been a constant traffic safety problem in certain types of roads. Although these crashes are not large in numbers, the outcomes are usually fatalities or severe injuries. Past studies on WWD crashes used either descriptive statistics or logistic regression to determine the impact of key contributing factors. In conventional statistics, failure to control the impact of all contributing variables on the probability of WWD crashes generates bias due to the rareness of these types of crashes. Distribution free methods, such as multiple correspondence analysis (MCA), overcome this issue, as there is no need of prior assumptions. This study used five years (2010-2014) of WWD crashes in Louisiana to determine the key associations between the contribution factors by using MCA. The findings showed that MCA helps in presenting a proximity map of the variable categories in a low dimensional plane. The outcomes of this study are sixteen significant clusters that include variable categories like determined several key factors like different locality types, roadways at dark with no lighting at night, roadways with no physical separations, roadways with higher posted speed, roadways with inadequate signage and markings, and older drivers. This study contains safety recommendations on targeted countermeasures to avoid different associated scenarios in WWD crashes. The findings will be helpful to the authorities to implement appropriate countermeasures.

Text Mining and Topic Modeling of Compendiums of Papers from Transportation Research Board Annual Meetings

The collective knowledge system has been advancing rapidly in the recent past. The digitalization of information in many online media—such as blogs, social media, articles, webpages, images, audios, and videos—provides an unprecedented opportunity for the extraction and identification of a knowledge trend. Prominent journal and conference proceedings usually contain extensive amounts of textual data that can be used to examine the research trends for various topics of interest and to understand how this research has helped in the advancement of a subject such as transportation engineering. The exploration of the unstructured contents in journal or conference papers requires sophisticated algorithms for knowledge extraction. This paper presents text mining techniques to analyze compendiums of papers published from TRB annual meetings, the largest and most comprehensive transportation conferences in the world. Topic models are algorithms designed to discover hidden thematic structure from massive collections of unstructured documents. This study used a popular topic model, latent Dirichlet allocation, to reveal research trends and interesting histories of the development of research by analyzing 15,357 compendiums of papers from 7 years (2008 to 2014) of TRB annual meetings.

Trends in Transportation Research: Exploring Content Analysis in Topics

Proceedings of journal and conference papers are good sources of big textual data to examine research trends in various branches of science. The contents, usually unstructured in nature, require fast machine-learning algorithms to be deciphered. Exploratory analysis through text mining usually provides the descriptive nature of the contents but lacks quantification of the topics and their correlations. Topic models are algorithms designed to discover the main theme or trend in massive collections of unstructured documents. Through the use of a structural topic model, an extension of latent Dirichlet allocation, this study introduced distinct topic models on the basis of the relative frequencies of the words used in the abstracts of 15,357 TRB compendium papers. With data from 7 years (2008 through 2014) of TRB annual meeting compendium papers, the 20 most dominant topics emerged from a bag of 4 million words. The findings of this study contributed to the understanding of topical trends in the complex and evolving field of transportation engineering research.

Using Deep Learning in Severity Analysis of At-Fault Motorcycle Rider Crashes

Motorcyclists are vulnerable highway users. Unlike passenger vehicle occupants, motorcycle riders do not have either protective structural surrounding or the advanced restraints that are mandatory safety features in cars and light trucks. Per vehicle mile traveled, motorcyclist fatalities occurred 27 times more frequently than passenger car occupant fatalities in traffic crashes. In addition, there were 4,976 motorcycle crash-related fatalities in the U.S. in 2014—more than twice the number of motorcycle rider fatalities that occurred in 1997. It shows that, in addition to current efforts, research needs to be conducted with additional resources and in newer directions. This paper investigated five years (2010–2014) of Louisiana at-fault motorcycle rider-involved crashes by using deep learning, which is a competent tool for mapping a high-multidimensional input into a smaller multidimensional output. The current study contributes to the existing injury severity modeling literature by developing a deep learning framework, named as DeepScooter, to predict motorcycle-involved crash severities. The final deep learning model can predict severity types with 100% accuracy with training data, and with 94% accuracy with test data, which is not attainable by using a statistical method or machine learning algorithm. The intensity of severities was found to be more likely associated with rider ejection, two-way roadways with no physical separation, curved aligned roadways, and weekends. It is anticipated that the DeepScooter framework and the findings will provide significant contributions to the area of motorcycle safety.

Vehicle Consumer Complaint Reports Involving Severe Incidents: Mining Large Contingency Tables

According to 2010–2014 Fatality Analysis Reporting System (FARS) data, nearly 6.35% of fatal crashes happened as a result of vehicles’ pre-existing manufacturing defects. The National Highway Traffic Safety Administration’s (NHTSA) vehicle complaint database incorporates more than 1.37 million complaint reports (as of June 1, 2017). These reports contain extended information on vehicle-related disruptions. Around 5% of these reports involve some level of injury or fatalities. This study had two principal objectives, namely (1) perform knowledge discovery to understand the latent trends in consumer complaints, and (2) identify clusters with high relative reporting ratios from a large contingency table of vehicle models and associated complaints. To accomplish these objectives, 67,201 detailed reports associated with injury or fatalities from the NHTSA vehicle complaint database were examined. Exploratory text mining and empirical Bayes (EB) data mining were performed. Additionally, this study analyzed five years (2010–2014) of FARS data to examine the research findings. Results show that major vehicular defects are associated with air bags, brake systems, seat belts, and speed controls. The EB metrics identified several key ‘vehicle model with major defect’ groups that require more attention. This study demonstrates the applicability of consumer complaints in identifying major vehicular defects as well as key groups of ‘vehicle model with major defect.’ The findings of this study will provide a significant contribution to the reduction of crashes from vehicle-related disruptions. The research presented in this paper is crucial given the ongoing advancement of connected and automated vehicle technologies.

Supervised association rules mining on pedestrian crashes in urban areas: identifying patterns for appropriate countermeasures

ABSTRACT In 2011, 4,432 pedestrians were killed (14% of total traffic crash fatalities), and 69,000 pedestrians were injured in vehicle-pedestrian crashes in the United States. Particularly in Louisiana, vehicle-pedestrian crashes have become a key concern because of the high percentage of fatalities in recent years. In 2012, pedestrians were accounted for 17% of all fatalities due to traffic crashes in Louisiana. Alcohol was involved in nearly 44% of these fatalities. This research utilized ‘a priori’ algorithm of supervised association mining technique to discover patterns from the vehicle-pedestrian crash database. By using association rules mining, this study aims to discover vehicle-pedestrian crash patterns using eight years of Louisiana crash data (2004–2011). The results indicated that roadway lighting at night helped in alleviating pedestrian crash severity. In addition, a few groups of interest were identified from this study: male pedestrians’ greater propensity towards severe and fatal crashes, younger female drivers (15–24) being more crash-prone than other age groups, vulnerable impaired pedestrians even on roadways with lighting at night, middle-aged male pedestrians (35–54) being inclined towards crash occurrence, and dominance of single vehicle crashes. Based on the recognized patterns, this study recommends several countermeasures to alleviate the safety concerns. The findings of this study will help traffic safety professionals in understanding significant patterns and relevant countermeasures to raise awareness and improvements for the potential decrease of pedestrian crashes.