Mining patterns of near-crash events with and without secondary tasks

The typical AV accident scenarios in the urban area obtained by clustering and association rule mining of real-world accident reports

Automated Vehicles (AVs) based on a collection of advanced technologies such as big data and artificial intelligence have opened an opportunity to reduce traffic accidents caused by human drivers. Nevertheless, traffic accidents of AVs continue to occur, which raises safety and reliability concerns about AVs. AVs are particularly vulnerable to accidents on urban roads than on highways due to various dynamic objects and more complex infrastructure. Several studies proposed a scenario-based approach of experimenting with the response of AVs to specific situations as a way to test their safety. Reliable and concrete scenarios are necessary to test AV safety under critical conditions accurately. This study aims to derive a typical accident scenario for evaluating the safety of AVs, specifically in urban areas, by analysing collisions reported by the DMV of California, USA. We applied a hierarchical clustering method to find groups of similar reports and then executed association rule mining on each cluster to correlate between accident factors and collision types. We combined statistically significant association rules to constitute a total of 14 scenarios that are described according to an adapted PEGASUS framework. The newly obtained scenarios exhibit significantly different accident patterns than the typical Human-driven Vehicles (HVs) in urban areas reported by National Highway Traffic Safety Administration. Our discovery urges AV safety to be tested reliably under scenarios more relevant than the existing HV accident scenarios.

Data mining approach to explore emergency vehicle crash patterns: A comparative study of crash severity in emergency and non-emergency response modes

Emergency vehicle crashes, involving police vehicles, ambulances, and fire trucks, pose a serious traffic safety concern causing severe injury and deaths to first responders and other road users. However, limited research is available focusing on the contributing factors and their interactions related to these crashes. This research aims to address this gap by 1) identifying patterns of emergency vehicle crashes based on severity levels in both emergency and non-emergency modes and 2) comparing the associations by response modes for the related fatal, nonfatal injury, and no-injury crashes. Two national crash databases, Fatality Analysis Reporting System (FARS) and Crash Report Sampling System (CRSS), were utilized for police-reported emergency vehicle crashes from January 2016 to February 2020. Association rule mining (ARM) was employed to reveal the association between factors that strongly contributed to these crashes. The generated rules were validated using the lift increase criterion (LIC). The results showed the complex nature of risk factors influencing the severity of emergency vehicle crashes. The fatal consequences of speeding with no seatbelt usage were evident for emergency mode, whereas none of these risky driving attributes was observed for non-emergency mode. In addition, the analysis identified the risk of fatal emergency vehicle crashes involving pedestrians in dark-lighted conditions in both response modes. Regarding nonfatal injury severity, angle collisions were more likely to occur at urban intersections during emergencies, while rear-end crashes were more frequent on segments with a posted speed limit of 40-45 mph during non-emergency incidents. The outcomes also revealed that the no-injury crashes involving fire trucks exhibited different patterns depending on the response mode. The findings of this study can guide in making effective strategies to improve safe driving behavior of first responders. The identified associations provide insights into the factors that can be controlled to ensure safe operation of emergency vehicles on the road.

On the impact of advanced driver assistance systems on driving distraction and risky behaviour: An empirical analysis of irish commercial drivers

Advanced driver assistance systems (ADAS) present promising benefits in mitigating road collisions. However, these benefits are limited when risky drivers continue engaging in distraction events. While there is evidence that real-time warnings help improve driving behaviour, the sustained benefits of warning-based ADAS on reducing driving distraction in light commercial vehicle (LCV) drivers remain unclear. This research determines the effect of receiving instant distraction warnings over two years using a naturalistic driving dataset comprising around one million trips from 373 LCV drivers in the Republic of Ireland. Furthermore, the study applies Association Rule Mining (ARM) to find the contextual variables (e.g., speed limit, road type, traffic conditions) that increase the likelihood of distraction events. The results show that warning-based ADAS providing real-time warnings helps reduce distraction events triggering driver inattention, forward collision, and lane departure warnings. Over half of the studied fleet reduced these warnings by at least 50% - lane departure after two months and driver inattention and forward collision after six months. It is found that both passive and active monitoring systems, coupled with coaching and rewards, significantly reduce aggressive driving behaviours tied to harsh acceleration (by 76%) and harsh braking (by 65%). The results of ARM show that the driving context introduces explanatory information for road safety programs. Low-speed urban roads and the summer season increase the likelihood of driver inattention and forward collision warnings. In contrast, high-speed rural roads increase the likelihood of lane departure warnings. These research findings support road safety stakeholders in developing risk assessments based on warning-based ADAS, targeted campaigns to reduce driving distraction, and driving coaching programs.

Risk Levels Classification of Near-Crashes in Naturalistic Driving Data

Identifying dangerous events from driving behavior data has become a vital challenge in intelligent transportation systems. In this study, we compared machine and deep learning-based methods for classifying the risk levels of near-crashes. A dataset was built for the study by considering variables related to naturalistic driving, temporal data, participants, and road geometry, among others. Hierarchical clustering was applied to categorize the near-crashes into several risk levels based on high-risk driving variables. The adaptive lasso variable model was adopted to reduce factors and select significant driving risk factors. In addition, several machine and deep learning models were used to compare near-crash classification performance by training the models and examining the model with testing data. The results showed that the deep learning models outperformed the machine learning and statistical models in terms of classification performance. The LSTM model achieved the highest performance in terms of all evaluation metrics compared with the state-of-the-art models (accuracy = 96%, recall = 0.93, precision = 0.88, and F1-measure = 0.91). The LSTM model can improve the classification accuracy and prediction of most near-crash events and reduce false near-crash classification. The finding of this study can benefit transportation safety in predicting and classifying driving risk. It can provide useful suggestions for reducing the incidence of critical events and forward road crashes.

Connections between Various Disorders: Combination Pattern Mining Using Apriori Algorithm Based on Diagnosis Information from Electronic Medical Records

Objective

Short-term or long-term connections between different diseases have not been fully acknowledged. This study was aimed at exploring the network association pattern between disorders that occurred in the same individual by using the association rule mining technique.

Methods

Raw data were extracted from the large-scale electronic medical record database of the affiliated hospital of Xuzhou Medical University. 1551732 pieces of diagnosis information from 144207 patients were collected from 2015 to 2020. Clinic diagnoses were categorized according to “International Classification of Diseases, 10th revision”. The Apriori algorithm was used to explore the association patterns among those diagnoses.

Results

12889 rules were generated after running the algorithm at first. After threshold filtering and manual examination, 110 disease combinations (support ≥ 0.001, confidence ≥ 60%, lift > 1) with strong association strength were obtained eventually. Association rules about the circulatory system and metabolic diseases accounted for a significant part of the results.

Conclusion

This research elucidated the network associations between disorders from different body systems in the same individual and demonstrated the usefulness of the Apriori algorithm in comorbidity or multimorbidity studies. The mined combinations will be helpful in improving prevention strategies, early identification of high-risk populations, and reducing mortality.