So far, studies for predicting construction safety accidents have mostly been conducted by statistical analysis methods that assume linear models, such as regression and time series analysis. However, it is difficult for this statistical analysis method to reflect the nonlinear characteristics of construction safety accidents determined by complex influencing factors. In general, deep learning techniques are used to analyze the nonlinear characteristics of complex influencing factors. Therefore, the purpose of this study is to propose a framework for developing a deep learning model for predicting safety accidents for sustainable construction. For this study, 1766 cases of actual accidents were collected by the Korea Occupational Safety Authority (KOSHA) over the 10-year period from 2010 to 2019. Eight factors influencing accident prediction such as medical day, progress rate, and construction scale were selected. Subsequently, the predictive power between deep learning models and conventional multi-regression models was compared using actual accident data at construction sites. As a result, a deep neural network (DNN) improved predictive power by 9.3% in mean absolute error (MAE) and 10.6% in root mean square error (RMSE) compared to a conventional multi-regression model. The results of this study provide guidelines for the introduction of deep learning technology to construction safety management.

Conventional statistical models, such as Poisson or negative binomial, have predefined underlying relationships between explanatory variables. However, artificial neural network (ANN), which overcomes the limitations of statistical prediction model, have gained popularity in practice and research for their ability to increase prediction accuracy. Thus, this study employs zero-truncated negative binomial (ZTNB) models and artificial neural network (ANN) models to analyze the distribution of railroad accident frequency and the corresponding number of casualties for 1995–2021 accident dataset of Korea’s national railroad. The study mainly focused on two most dominant accident types which are human-involved accidents (accounted for 89.2% of all accidents) and ground-level crossing accidents (9.6%) from the historic dataset. This is because not just data proportion, rather such accident types were received very little attention compared to fatal train accident in the accident prediction study. Further, these types of accident showed clearly tended to decrease over time, but time trend has been found very weak at the type of fatal train accident. The performance of the developed models was estimated by mean square error (MSE), the root mean square error (RMSE), and the coefficient of determination (R²). Results present that ANN models outperform ZTNB models in fitting and prediction, demonstrating once again ANN’s superiority over statistical models for predicting accident frequency and casualty count.

Railroads play a pivotal role in the Korean national economy, necessitating a thorough understanding of factors influencing accidents for effective mitigation strategies. Unlike prior research focused on accident frequency and severity, this study delves into the often-overlooked aspect of time delays resulting from railroad accidents. Analyzing 15 years of nationwide data (2008–2022), encompassing 3244 human-related and 3350 technical events, this research identifies key factors influencing delay likelihood and duration. Factors considered include event type, season, train type, location, operator size, person type involved, facility type, and causes. Despite an overall decrease in events, variable delay times highlight the need to comprehend specific contributing factors. To address excess zeros, the study employs a two-stage model and a zero-inflated negative binomial (ZINB) model, alongside artificial neural networks (ANNs) for non-linear pattern recognition. Human-related delays are influenced by event types, seasons, and passenger categories, exhibit nuanced impacts. Technical-related delays are influenced by incident types and facility involvement. Regarding model performance, the ANN models outperform regression-based models consistently in all cases. This study emphasizes the importance of considering both human and technical factors in predicting and understanding railroad accident delays, offering valuable insights for formulating strategies to mitigate service disruptions associated with these incidents.

The purpose of this research is to build a deep learning algorithm-based model that can use weather indicators to quantitatively predict financial losses associated with weather-related railroad accidents. Extreme weather events and weather disasters caused by global warming are happening with increasing frequency worldwide, leading to substantial economic losses. Railways, which represent one of the most important means of transportation, are also affected by such weather events. However, empirical and quantitative studies examining losses stemming from weather conditions for railways have to this point been scarce. Hence, the present study collected and analyzed weather-induced railway accident data and meteorological factors (wind, precipitation, rainfall, etc.) from 2001 to 2021 with the aim of predicting financial losses caused by weather events; the ultimate goal is to help inform long-term strategies for effective recovery from railway accidents. Objective and scientific analysis was conducted in the present study by using a deep learning algorithm. The outcomes and framework of this research will offer crucial guidelines for efficient and sustainable railway maintenance. These results will also serve as a crucial point of reference for loss quantification studies and other facility management studies.

Household members often interact with each other during their daily activity- and travel-related decision-making process. In the context of social and recreational activities, these interactions lead to decisions about pursuing such episodes either independently or jointly with others. The substantial focus of the study is on the operationalizing household interactions in various modeling mechanisms by the ways of seeking to maximize the utility of each household member in an interdependent fashion. Florida add-on samples from the 2009 National Household Travel Survey are utilized. The results indicate that household income, age of male, engagement of mandatory activities, engagement of maintenance activities, day of the week, and whether raining have strong impacts on the choice of social-recreational activity participations. Overall, no one model turned out to be a clear “winner” in terms of statistical goodness of fit and predictive abilities, rather on the behavioral insights, the multi-linear logit and parallel constrained logit models do highlight differences in the “power” across household heads, and the trivariate binary probitmodels capture strong correlations in the choices across the male and female and between solo and joint choices.

Predicting the vehicle used for individual trips and tours can help improve the quantification of energy consumption, quality of emissions forecasts, and assess impacts of policies that vary by vehicle type. The focus of this study is to contribute to incorporating this aspect within current activity-based modeling frameworks. Data from the 2009 National Household Travel Survey of the United States are used in this analysis. The empirical scope of this study is limited to two-adult two car households in the context of social-recreational tours; however, this methodology can be directly extended to other cases as well. An exploratory analysis indicates that the “primary driver” variable is perhaps the strongest predictor of the vehicle allocated to independent tours and tours made by adults with children. In the case of joint tours, there is clearly a choice of vehicle to be made. Following the exploratory analysis, two models were developed. One allocates each vehicle to a primary driver in the household (long-term, household-level model). The second allocates a vehicle for the joint tours (short-term, tour-level model). Both models were estimated using the unlabeled binary-logit methodology. Several vehicle attributes (such as size/body type, fuel efficiency, age, and operating costs) and socio-economic variables (age, and presence of children) were estimated to be statistically-significant predicators of the vehicle-allocation patterns.