survival analysis based on the aft approach with a semi-continuous response under zero-inflation: financial applications

In actuarial science, mortality models serve as essential tools that utilize historical information to forecast future patterns and compute mortality rates, particularly for life insurance and premium assessments. these models necessitate sufficient flexibility to accommodate various factors, including age and product categories. time-to-event (tte) data, which encompasses non-negative, often asymmetric durations and censoring, is crucial in survival analysis. traditional regression approaches are not suitable for censored data, necessitating specialized methods such as cox proportional hazards and accelerated failure time (aft) models. while cox models assume proportional hazards, aft models, which can incorporate various distributions, provide more readily interpretable results in complex situations. this research enhances aft models by introducing a zero-inflation component to address data concentrations near zero, effectively transforming continuous distributions into semi-continuous ones. by combining zero-inflation with right or left censoring, the study introduces novel models that expand the applicability of survival analysis in actuarial contexts.