Occupational risk assessment is the process of estimating the magnitude of risks that cannot be avoided. Then, the corresponding assessment is carried out, using comparative tables with different evaluation methods. Current risk assessment techniques enable the individual assessment of each potential risk, but there is no method to globally assess potential risks in an organization. The motivation of this research was to develop an objective and quantitative risk assessment system through a diffuse probabilistic model integrating stochastic and non-stochastic uncertainty. To this effect, an empirical collective record was used, whose attribute of interest was the occurrence of different accident types over a period of 52 weeks. Here, each of the collectives represented a linguistic input variable. In the probabilistic fuzzification stage, the frequentist probability of the occurrence of accidents was determined. One of our most important contributions to probabilistic fuzzy systems lies in our classification of language labels based on the linguistic projection of frequentist probabilities via a projection membership function determined by experts. The use of the total probability theorem in the implication stage is also proposed. The output of the system determines the type of risk, its evaluation, and the probability of its occurrence, vital factors to be considered in prevention work. The system’s stages are explicitly described and applied to real data corresponding to construction materials distribution company. One of the relevant conclusions of this research is that the integration of stochastic and imprecise uncertainty allows for a more reliable risk assessment system.