Abstract: In terms of the risk assessment problem with the lack of information, this paper constructs a novel risk assessment model based on linguistic ELECTRE-TRI in consideration of the heterogeneity and non-compensation among the criteria. Firstly, the experts are asked to characterize and handle the uncertain information in the utilization of CFGJ or its extended format, PD-HFLTS. Secondly, the outranking relations between risk assessment values of potential dangerous events and the threshold value of risk grade division are constructed with the utilization of ELECTRE-TRI. And then the risk levels of the potential dangerous events can be attained on the basis of the outranking relations obtained before. Finally, an example validates the scientific and effective nature of the model. The results show that CFGJ and PD-HFLTS can not only effectively characterize the uncertainty in risk assessment information, but also improve the accuracy of the process of semantic information, and that the proposed model based on ELECTRE-TRI has the outstanding feasibility and effectiveness in addressing the multi-criteria risk assessment problems with the non-compensation.

Key words: risk assessment, context-free grammar judgement(CFGJ), non-compensation, uncertainty, possibility distribution-hesitant fuzzy linguistic term set(PD-HFLTS), elimination and choice expressing reality tree(ELECTRE-TRI)