Supervised Learning : Exploring Parametric Accelerated Failure Time Models for Estimating Lifetimes in Survival Data
This project implements Accelerated Failure Time Models based on the Weibull, Log-Normal and Log-Logistic distributions to analyze time-to-event data by directly modelling the survival time. The statistically significant model predictors were identified with their effects on the estimated lifetime evaluated using their acceleration factors. The resulting predictions derived from the candidate models were evaluated in terms of their discrimination power, fit and calibration performance using the concordance index metric, mean absolute error and brier score. Feature impact on model output were assessed using Shapley Additive Explanations. Survival and hazard functions were estimated.
