Predicting Human Error in Industrial Operation with EEG and Data Mining Techniques
MetadataShow full item record
Human error was a factor in many accidents, including the Bhopal pesticide plant explosion, Hillsborough football stadium disaster, Paddington and Southall rail crashes, capsizing of the Herald of Free Enterprise, Chernobyl and Three-Mile Island incidents. It has been estimated that human error is one of causes for up to 90% of all workplace accidents. Due to potential severe damage and economic loss of accidents caused by human error, it is import to proactively prevent the accidents before happening. Existing research mostly focused on improving operation interfaces for reducing worker’s erroneous operations. However, it is hard to say whether these methods can prevent erroneous operation. This research is to build up a system which can predict the errors in advance and then avoid the occurrence of accidents and economic loss. By using electroencephalography (EEG) technology, the LDA method in data mining is able to differentiate different human brain wave characteristics and patterns 200 ms before human error actually happened. The experiment result shows that the average d’ of the prediction result in the best case is 1.57, average hit rate is 0.57, the false alarm rate is 0.29, and the average area size under curve is 0.59 for each feature point. Further development of the system in preventing human errors in workplaces is discussed.