Approaches to Accounting for the Uncertainty of Data on the Time of Realization of Episodes in Models for Assessing the Characteristics of Individuals Episodic Behavior

The gamma Poisson model of behavior arises in the context of analyzing patterns of episodic behavior of individuals and is intended for building the estimates of behavior characteristics with data on the episodes of behavior. Such problems arise in the investigation of risk-related behavior. However, if the data collected are self-reported, then it may be imprecise due to recall bias or natural language formulation. The paper presents approaches for managing emerging uncertainty in models for estimation of the behavior characteristics: regression and hybrid Bayesian network of trust. Computational experiments have been carried out and the stability of existing models is analyzed.