Modeling and predicting the changes in hearing loss of workers using neural network data mining algorithm: field study

Background: one of the most common physical harm factors in the workplace is noise. The most important and definitive harmful effects of noise include hearing loss both permanent and temporary. Globally, 12% of people are subject to noise-induced hearing loss (NIHL). This study was conducted with the aim of modeling and predicting the changes in the hearing loss of workers using neural network data mining algorithm.Materials & Methods: this cross-sectional descriptive study of prospective analytical type was conducted in 2018 in a mine industry in the south west of Iran. In this study, the workers were categorized into three groups of exposure to different noise pressure levels (one control group and two case groups). In each group 50 subjects and overall three audiometric groups were performed for 150 of workers. The stages included: 1. Investigating the demographic information of the studied individuals 2. Selecting variables (predictors) to model the hearing loss 3.Performing the audiometry for both ears 4.Calculating permanent threshold shift for both left and right ears separately 5.Calculating permanent threshold shift of both ears 6.Classifying different types of hearing loss 7. Modeling changes in hearing loss 8. Determining the rate of error and accuracy of the model. For modeling and predicting the changes in the hearing loss of workers, neural network data mining algorithm was used. SPSS 18 was employed to analyze the statistical tests including linear regression and paired t-test. Also, IBM SPSS Modeler 18.0 software was employed to model the neural network algorithm.Results: the results indicated that in the first model (SPL<70 dBA), 8 and 1 kHz frequencies with the weights of 38% and 6% had the maximum and minimum effects, respectively. In the second model (SPL 70-80 dBA), 4KHz and age factors had the maximum and minimum effects with the weights of 19% and 7%, respectively. In the third model (SPL> 85 dBA), 4KHz and work background had the maximum and minimum effects with the weights of 20 and 6%, respectively. In the fourth model, 4KHz and 2 KHz with the weights of 18 and 3% had the maximum and minimum effects, respectively. In the fifth model, 4KHz and age with the weights of 18% and 1% had the maximum and minimum effects, respectively. Conclusion: the modeling of changes in hearing loss by the neural network algorithm predicted the high impact of 4KHz frequency in the variations of hearing loss. Based on the high accuracy obtained in this modeling, this algorithm is a suitable and powerful tool for predicting and modeling hearing loss.

کلیدواژه ها:

noise ، modeling ، hearing loss ، data mining ، neural network algorithm