A Study on Data Pre-Processing and Accident Prediction Modelling for Occupational Accident Analysis in the Construction Industry

Abstract

In the construction industry, it is difficult to predict occupational accidents because various accident characteristics arise simultaneously and organically in different types of work. Furthermore, even when analyzing occupational accident data, it is difficult to deduce meaningful results because the data recorded by the incident investigator are qualitative and include a wide variety of data types and categories. Recently, numerous studies have used machine learning to analyze the correlations in such complex construction accident data; however, heretofore the focus has been on predicting severity with various variables, and several limitations remain when deriving the correlations between features from various variables. Thus, this paper proposes a data processing procedure that can efficiently manipulate accident data using optimal machine learning techniques and derive and systematize meaningful variables to rationally approach such complex problems. In particular, among the various variables, the most influential variables are derived through methods such as clustering, chi-square, Cramer’s V, and predictor importance; then, the analysis is simplified by optimally grouping the variables. For accident data with optimal variables and elements, a predictive model is constructed between variables, using a support vector machine and decision-tree-based ensemble; then, the correlation between the dependent and independent variables is analyzed through an alluvial flow diagram for several cases. Therefore, a new processing procedure has been introduced in data preprocessing and accident prediction modelling to overcome difficulties from complex and diverse construction occupational accident data, and effective accident prevention is possible by deriving correlations of construction accidents using this process.

Determining the Short-term Susceptibility of Construction Workers to Occupational Accidents Using Stochastic Gradient Boosting

“…First, despite the applicability of many ML methods such as KNN [25], NB [17], RF [10], SVM [11], and ANN [9] has been evaluated in the construction safety management literature, the use of SGB method is examined in this research. Besides, despite some researhers considering during-or post-accident attributes to predict several output variables in the construction safety management research field [27,31], the current study eliminated this limitation by only considering the pre-accident attributes. This approach is particularly beneficial as proactive safety management framework can only be developed by incorporating pre-accident inputs.…”

“…Ajayi et al [16] adopted deep neural network approach for the prediction of the occurrence of accidents, injured body part, lost time, and damage due to accidents. Lee et al [27] considered seven output variables such as year of the accident, type of work, assailing materials, severity of accident, body part affected, accident cause, and accident type, and each was predicted by taking other six as input variables. Comprehensive literature review shows that there is a variety of research in the subject matter that adopted ML methods.…”

Prediction model of seasonality in the construction industry based on the accidentality phenomenon

“…Analysis of the accidentality phenomenon in the construction industry can also be found in [24][25][26][27]. The research conducted by Hoła [28] regarding the Polish construction industry indicated that the accident rate in the construction industry changes every year, and that there is a clear downward trend.…”

Prediction of construction accident outcomes based on an imbalanced dataset through integrated resampling techniques and machine learning methods

“…Data collection for providing a basis to predict occupational accidents is important to attain accurate results (Ayhan and Tokdemir, 2019). Therefore, several researchers proposed different prediction/classification models based on the historical data collected from various institutions or companies (Ajayi et al, 2020;Alexander et al, 2017;Baker et al, 2020a;Choi et al, 2020;Lee et al, 2020). However, a dataset consisting of fatal and nonfatal construction accidents can be severely imbalanced (Baker et al, 2020b;Choi et al, 2020).…”

“…The study also highlights a city-basis investigation, which analyzes 121,002 construction accidents occurred in Istanbul. Past predictive studies on construction safety mainly rely on accident-basis (Jahangiri et al, 2019;Yang et al, 2016), project type-basis (Ajayi et al, 2020), company-basis (Lee et al, 2020) or country-basis (Choi et al, 2020;Kang and Ryu, 2019) implementations. However, metropolitan cities, such as Istanbul, have their own sector dynamics with the number and variety of projects.…”