Variance-based global sensitivity analysis for rear-end crash investigation using deep learning

Analyzing speed-difference impact on freeway joint injury severities of Leading-Following vehicles using statistical and data-driven models

Rear-end (RE) crashes are notably prevalent and pose a substantial risk on freeways. This paper explores the correlation between speed difference among the following and leading vehicles (Δν) and RE crash risk. Three joint models, comprising both uncorrelated and correlated joint random-parameters bivariate probit (RPBP) approaches (statistical methods) and a cross-stitch multilayer perceptron (CS-MLP) network (a data-driven method), were estimated and compared against three separate models: Support Vector Machines (SVM), eXtreme Gradient Boosting (XGBoost), and MLP networks (all data-driven methods). Data on 15,980 two-vehicle RE crashes were collected over a two-year period, from January 1, 2021, to December 31, 2022, considering two possible levels of injury severity: no injury and injury/fatality for both drivers of following and leading vehicles. The comparative performance analysis demonstrates the superior predictive capability of the CS-MLP network over the uncorrelated/correlated joint RPBP model, SVM, XGBoost, and MLP networks in terms of recall, F-1 Score, and AUC. Significantly, numerous shared variables influence the injury severity outcomes for the following and leading vehicles across both statistical and data-driven approaches. Among these factors, the following vehicle (a truck) and the leading vehicle (a passenger car) demonstrate contrasting effects on the injury severity outcomes for both vehicles. Furthermore, the SHapley Additive exPlanations (SHAP) values from the CS-MLP network visually show the relationship between Δν and injury severity, revealing non-linear trends unlike the average effects shown by statistical methods. They indicate that the least injury outcomes for both following and leading vehicles occurs at a Δν of 0 to 10 mph, matching observed patterns in RE crash data. Additionally, a marked variation in the trend of SHAP values for the two vehicles is noted as the speed difference increases. Therefore, the findings affirm the superior performance of joint model development and substantiate the non-linear impacts of speed difference on injury outcomes. The adoption of dynamic speed control measures is recommended to mitigate the injury outcomes involved in two-vehicle RE crashes.

An investigation of heterogeneous impact, temporal stability, and aggregate shift in factors affecting the driver injury severity in single-vehicle rollover crashes

Single-vehicle rollover crashes have been acknowledged as a predominant highway crash type resulting in serious casualties. To investigate the heterogeneous impact of factors determining different injury severity levels in single-vehicle rollover crashes, the random parameters logit model with unobserved heterogeneity in means and variances was employed in this paper. A five-year dataset on single-vehicle rollover crashes, gathered in California from January 1, 2013, to December 31, 2017, was utilized. Driver injury severities that were determined to be outcome variables include no injury, minor injury, and severe injury. Characteristics pertaining to the crash, driver, temporal, vehicle, roadway, and environment were acknowledged as potential determinants. The results showed that the gender indicator specified to minor injury was consistently identified as a significant random parameter in four years' models and the joint five-year model, excluding the 2016 crash model where the night indicator associated with no injury was observed to produce the random effect. Additionally, two series of likelihood ratio tests were conducted to assess the year-to-year and aggregate-to-component temporal stability of model estimation results. Marginal effects of explanatory variables were also calculated and compared to analyze the temporal stability and interpret the results. The findings revealed an overall temporal instability of model specifications across individual years, while there is no significant aggregate-to-component variation. Injury severities were observed to be stably affected by several variables, including improper turn indicator, under the influence of alcohol indicator, old driver indicator, seatbelt indicator, insurance indicator, and airbag indicator. Furthermore, the year-to-year and aggregate-to-component shift was quantified and characterized by calculating the differences in probabilities between within-sample observations and out-of-sample predictions. The overall results imply that continuing to expand and refine the model to incorporate more comprehensive datasets can result in more robust and stable injury severity prediction, thus benefiting in mitigating the associated driver injury severity.

CLDTLog: System Log Anomaly Detection Method Based on Contrastive Learning and Dual Objective Tasks

System logs are a crucial component of system maintainability, as they record the status of the system and essential events for troubleshooting and maintenance when necessary. Therefore, anomaly detection of system logs is crucial. Recent research has focused on extracting semantic information from unstructured log messages for log anomaly detection tasks. Since BERT models work well in natural language processing, this paper proposes an approach called CLDTLog, which introduces contrastive learning and dual-objective tasks in a BERT pre-trained model and performs anomaly detection on system logs through a fully connected layer. This approach does not require log parsing and thus can avoid the uncertainty caused by log parsing. We trained the CLDTLog model on two log datasets (HDFS and BGL) and achieved F1 scores of 0.9971 and 0.9999 on the HDFS and BGL datasets, respectively, which performed better than all known methods. In addition, when using only 1% of the BGL dataset as training data, CLDTLog still achieves an F1 score of 0.9993, showing excellent generalization performance with a significant reduction of the training cost.

Applications of artificial intelligence in forensic sciences: Current potential benefits, limitations and perspectives

In recent years, new studies based on artificial intelligence (AI) have been conducted in the forensic field, posing new challenges and demonstrating the advantages and disadvantages of using AI methodologies to solve forensic well-known problems. Specifically, AI technology has tried to overcome the human subjective bias limitations of the traditional approach of the forensic sciences, which include sex prediction and age estimation from morphometric measurements in forensic anthropology or evaluating the third molar stage of development in forensic odontology. Likewise, AI has been studied as an assisting tool in forensic pathology for a quick and easy identification of the taxonomy of diatoms. The present systematic review follows the PRISMA 2020 statements and aims to explore an emerging topic that has been poorly analyzed in the forensic literature. Benefits, limitations, and forensic implications concerning AI are therefore highlighted, by providing an extensive critical review of its current applications on forensic sciences as well as its future directions. Results are divided into 5 subsections which included forensic anthropology, forensic odontology, forensic pathology, forensic genetics, and other forensic branches. The discussion offers a useful instrument to investigate the potential benefits of AI in the forensic fields as well as to point out the existing open questions and issues concerning its application on real-life scenarios. Procedural notes and technical aspects are also provided to the readers.