Loretta Bortey

Despite growing interest in the application of machine learning (ML) for accident prediction and safety analysis, limited research has explored its use in predicting anatomical injury risk among highway workers. This study addresses this gap by developing a predictive model capable of classifying body parts most susceptible to injury in highway-related incidents. Positivism and interpretivism set the theoretical foundations for this study. The sequential exploratory mixed method adopted involved the preprocessing of accident datasets, feature selection and model evaluation using established performance metrics. A Support Vector Machine (SVM) algorithm was employed as the primary classifier, with its performance benchmarked against three comparative models: Naïve Bayes (NB), Random Forest (RF) and a Recurrent Neural Network (RNN). Analysis results showed that variables such as ‘region’, ‘site/project’, ‘event type’, ‘vehicles involved’ and ‘location’ were very significant in predicting bodily injuries. Moreover, the findings also indicate that the SVM model, when optimally tuned, yields competitive classification accuracy, with RF and RNN models showing promising supplementary performance. This study introduces a novel framework for body-part injury classification within high-risk highway environments tailored for highway workers. This is the first study to use real life datasets specifically collected from highway worker injuries and departs from previous studies which have focused on drivers, pedestrians and the road only.

Highway traffic officers (HTOs) are often exposed to life-threatening workplace incidents while performing their duties. However, scant research has been undertaken to address these safety concerns. This research explores case study data from highway incident reports (held by National Highways, a UK government company) and employs deep neural network (DNN) in unearthing patterns which inform safety decision makers on pertinent safety challenges confronting HTOs. A mixed philosophical stance of positivism and interpretivism was adopted to synthesise the findings made. A four-phase sequential method was implemented to evaluate the validity of the research viz.: (i) architectural design; (ii) data exploration; (iii) predictive modelling; and (iv) performance evaluation. The DNN model’s predictive performance is benchmarked against three other machine learning models, namely Support Vector Machines (SVM), Random Forest (RF), and Naïve Bayes (NB). The DNN model outperformed the other three models. Findings from the data exploration also show that most work operations undertaken by HTOs have a medium risk level with night shifts posing the greatest risk challenges. Carriageways and traffic management enclosures had the highest incident occurrence. This is the first study to uncover such hidden patterns and predict risk levels using a database specifically for HTOs. This study presents evidence-based information for proactive risk management for HTOs.

This study conducts a systematic review of safety risk models and theories by summarizing and comparing them to identify the best strategies that can be adopted in a digital ‘conceptual’ safety risk model for highway workers’ safety. A mixed philosophical paradigm was adopted (that used both interpretivism and post-positivism couched within inductive reasoning) for a systematic review and comparative analysis of existing risk models and theories. The underlying research question formulated was: can existing models and theories of safety risk be used to develop this proposed digital risk model? In total, 607 papers (where each constituted a unit of analysis and secondary data source) were retrieved from Scopus and analysed through colour coding, classification and scientometric analysis using VOSViewer and Microsoft Excel software. The reviewed models were built on earlier safety risk models with minor upgrades. However, human elements (human errors, human risky behaviour and untrained staff) remained a constant characteristic, which contributed to safety risk occurrences in current and future trends of safety risk. Therefore, more proactive indicators such as risk perception, safety climate, and safety culture have been included in contemporary safety risk models and theories to address the human contribution to safety risk events. Highway construction safety risk literature is scant, and consequently, comprehensive risk prevention models have not been well examined in this area. Premised upon a rich synthesis of secondary data, a conceptual model was recommended, which proposes infusing machine learning predictive models (augmented with inherent resilient capabilities) to enable models to adapt and recover in an event of inevitable predicted risk incident (referred to as the resilient predictive model). This paper presents a novel resilient predictive safety risk conceptual model that employs machine learning algorithms to enhance the prevention of safety risk in the highway construction industry. Such a digital model contains adaptability and recovery mechanisms to adjust and bounce back when predicted safety risks are unavoidable. This will help prevent unfortunate events in time and control the impact of predicted safety risks that cannot be prevented.