Classifying Workers into Risk Sensibility Profiles: A Neural Network Approach

In this paper we propose a neural network-based classifier to associate a worker with his/her risk sensibility profile. The basic idea behind the risk sensibility profile is that risks are preventable by appropriate actions that decrease their injurious potential. Also, some criticality factors have been shown to be connected with risk perception and risk propensity. Mapping workers into risk sensibility profiles means to measure how safely workers interact with the risks they are exposed to, by considering the preventing actions they perform, and their criticality factors. The main advantages of the proposed classification consist in: (i) supporting the selection of the most suitable worker to safely perform a given task, (ii) tailoring the safety training to each worker's need, to effectively decrease the probability of injury. The proposed neural classifier was trained by using interviews we collected within some volunteer shoe factories. Workers were asked to indicate the preventive actions they would perform if exposed to one or more risks, among a set of proposed actions. Also, workers answered questions to associate a value with each criticality factor. Two typical tasks of the footwear industry, characterized by one and two risks, respectively, were considered to validate and test the classifier.