Predicting Sequences of Progressive Events Times with Time-dependent Covariates

TL;DR

This paper introduces a novel modeling approach for predicting sequences of progressive events using time-dependent covariates, focusing on state indicator processes rather than hazard functions, applicable in medical and agricultural contexts.

Contribution

It proposes a new method that models state indicator processes for event prediction, avoiding direct hazard function modeling and enabling easy incorporation of time-dependent covariates.

Findings

Effective in predicting event sequences in real-world data

Flexible application to medical and agricultural problems

Improves prediction accuracy with time-dependent covariates

Abstract

This paper presents an approach to modeling progressive event-history data when the overall objective is prediction based on time-dependent covariates. This approach does not model the hazard function directly. Instead, it models the process of the state indicators of the event history so that the time-dependent covariates can be incorporated and predictors of the future events easily formulated. Our model can be applied to a range of real-world problems in medical and agricultural science.