Development of a Prediction Model of Blood Lead Levels among Employees in a Metal Soldering Factory by Using Artificial Neural Networks

Abstract

The objectives of this research were: 1) to investigate the blood lead levels of workers in a metal soldering factory from their annual health check-up; and 2) to develop a prediction model of blood lead levels among the workers by using artificial neural networks. The instruments in this study consisted of the data sets for developing an appropriate prediction model of blood lead levels by using artificial neural networks, and for predicting the blood lead levels This study started with the analysis of the blood lead levels of 46 workers in a metal soldering factory in Samut Prakarn Province. The workers had 10 times of annual health check-up results during the years 2008 to 2012. Blood lead levels of the workers were predicted using data input of the artificial neural networks including work experience, nature of work, health check-up results, blood hemoglobin concentration, amount of raw materials (lead and tin), and investigating results of stack and workplace air lead levels. Percentage, Pearson Correlation Coefficient, Magnitude of Relative Error, Mean Magnitude of Relative Error, and accuracy were utilized in the data analysis. The results indicated that: 1) the blood lead levels of all 46 workers from annual health check-up were less than 40 micrograms per cubic meter. Most workers had amount of blood lead 6 -10 micrograms per cubic meter (50.4%), followed by 11 - 15 micrograms per cubic meter (32.2 %); and 2) the model was suitable to predict blood lead levels with prediction error of 1.44 %. Prediction of blood lead levels of the workers by using the artificial neural networks had accuracy up to 90.17%.