Method of fault prediction for avionics components based on stacking regression

Abstract As avionics systems become increasingly complex, traditional fault prediction methods are no longer sufficient to meet modern demands. This paper introduces four advanced fault prediction methods for avionics components, utilising a multi-step prediction strategy combined with a stacking regressor. By selecting various standard regression models as base regressors, these base regressors are first trained on the original data, and their predictions are subsequently used as input features for training a meta-regressor. Additionally, the Tree-structured Parzen Estimator (TPE) algorithm is employed for hyperparameter optimisation. The experimental results demonstrate that the proposed stacking regression methods exhibit superior accuracy in fault prediction compared to traditional single-model approaches.