Belief function theory on the continuous space with an application to model based classification

The paper defines belief functions on continuous frames of discernment, where masses generalize into densities. Explicit and manageable solutions can be formulated when densities are only assigned to the intervals of R. When our domain knowledge is represented by the pignistic probability density, then we build the corresponding least committed belief function. The theory is applied to model based classification and the results are compared to the classical Bayesian approach.