An Adaptable and Intelligent Railway Transportation System with Novel Ensemble Machine Learning Architecture

Railway systems are constantly evolving to keep up with the increasing demands, requiring greater reliability and safety. In this research, novel machine learning architectures have been proposed to address the challenges of accurate failure prediction. The proposed methods include, the Time-weighted Ensemble LSTM architecture which exploits the time series representation of train journeys. Secondly, an LSTM based encoder-decoder surrogate model for hyperparameter optimisation. Leveraging the properties of decision trees and recurrent neural networks to improve adaptability and interpretability, a novel Bivariate-split decision tree – LSTM ensemble is proposed.