Artificial Neural Network for Predicting Silicosis Risk Scores among Stone Carvers in Nakhon Ratchasima, Thailand

Abstract

Working in an environment contaminated with silica dust is increasing the risk for the development silicosis for stone carvers in Nakhon Ratchasima province, Thailand. This study aimed to develop an Artificial Neural Network (ANN) model to predict silicosis risk scores among 243 stone carvers who exposed to silica at work from August and October 2023 in Nakhon Ratchasima, Thailand. Regression analysis was performed in order to find the factors significantly associated with silicosis risk scores. Only 4 influenced variables were tested by using regression analysis. Regression analysis and ANN were run to predict silicosis risk scores from 4 influential variables. These influential variables included the concentration of silica dust exposure (mg/m3), working hours per day ( hour) , congenital disorder, and separation of residence from a workplace. ANN model was constructed as 4- 3- 2- 1 by comprising of 4 input variables, 3 and 2 hidden nodes, 1 output variable, momentum was 0.05, learning rate was 0.5, and learning time was 100,000 epochs. The findings showed that the least error with the Mean Absolute Percentage Error (MAPE) was 4.58. Predictive accuracy was assessed by MAPE value. ANN model showed the least value of MAPE when comparing an error value of less than 10%. Therefore, the ANN model is accurate and valid for silicosis risk scores prediction in individuals in order to plan for solving problems according to the factors influencing the silicosis risk scores before starting to work. Further research is recommended to improve to model by large-sample-size research.