Classification of jobs with risk of low back disorders by applying data mining techniques

Abstract

Work related low back disorders (LBDs) continue to pose significant occupational health problem that affects the quality of life of the industrial population. The main objective of this study was to explore the application of various data mining techniques, including neural networks, logistic regression, decision trees, memory-based reasoning, and the ensemble model, for classification of industrial jobs with respect to the risk of work-related LBDs. The results from extensive computer simulations using a 10-fold cross validation showed that memory-based reasoning and ensemble models were the best in the overall classification accuracy. The decision tree and memory-based reasoning models were the most accurate in classifying jobs with high risk of LBDs, whereas neural networks and logistic regression were the best in classifying jobs with low risk of LBDs. The decision tree model delivered the most stable results across 10 generations of different data sets randomly chosen for training, validation, and testing. The classification results generated by the decision tree were the easiest to interpret because they were given in the form of simple 'if-then' rules. These results produced by the decision tree method showed that the peak moment had the highest predictive power of LBDs.