A Neuro-fuzzy approach to identify a hierarchical fuzzy system for modelling aviation pilot attention

Attention has been shown to be a predictor of flight performance and, therefore, it is necessary to assess this cognitive ability to evaluate candidate aviation pilots and to verify if the pilot has the sufficient attention level required for the flight duties. In this paper, we present a study that uses a Neuro-fuzzy approach to identify a benchmark model of the aviation pilot attention level. The model is learned from the data examples collected using a computerized battery of seven tests, which was specifically built and validated to assess the main cognitive factors related with the aviation pilot attention in realistic scenarios. Data examples were collected in experimental session with a total of 180 participants, 96 military aviation pilots and 84 untrained people as controls. The aim is to build the model as a classifier that is able to discriminate between the two groups, allowing to identify the peculiar profile of the aviation pilot as opposed to control subjects. Classification performance analysis shows that a hierarchical fuzzy system has a better accuracy than single stage classification algorithms and gives more details about the different attention factors. Moreover, a fuzzy system for our model because it can be readable by human instructors and used as guidance for the training. The ultimate objective of our work is an expert system that will be able to assess the attention performance and compare it against the typical profile of the aviation pilot. This will be used as an instrument for more accurate selection and training of aviation pilots by identifying the areas of deficit that need to be improved, and to measure if a pilot has a sufficient level of attention before the flight.