A Model to Predict Occupational Safety and Health Management Expenses in Construction Applying Multi-variate Regression Analysis and Deep Neural Network

To reduce safety accidents leading to serious casualties and compensation, the Ministry of Employment and Labor prescribes Occupational Safety and Health Management Expenses (OSHME). Though there is an expense calculated by fixed rate, it is more urgent to spend the set amount according to the situation rather than standards due to the ambiguous criteria. Consequently, OSHME used nominally to make contract rather than to educate and protect the safety of workers. Therefore, in this research, OSHME was predicted by applying Deep Neural Network (DNN) with various optimizer, epoch, nodes based on 135 general construction cases under 500 million won to compare from multivariate regression analysis and origin contract cost multiplied existing rate by applying error indicators, mean squared error (MSE) and mean absoloute error (MAE). As a result, by comparing the values from three different analysis, DNN model with bayesian regularization optimizer in 0.01 learning rate was outstanding method to predict OSHME. Rather than simply executing as the current law, multiplying direct labor and material costs by a certain percentage, proposed model would support to calculate construction costs efficiently. Especially, as the contract material costs show high impact on consumed OSHME, when the sum of labor and material costs is the same, if material costs are high, it is required that OSHME be set higher. Furthermore, it is necessary to specify clear criteria and detailed usage plans to ensure not to execute incorrectly.