PMTT: Parallel multi-scale temporal convolution network and transformer for predicting the time to aging failure of software systems

Kai Jia, Xiao Yu, Chen Zhang, Wenzhi Xie, Dongdong Zhao, Jianwen Xiang

Abstract: Highlights•A TTAF prediction framework with PMTT is proposed to extract degradation information from the software systems.•Multiple sets of run-to-failure data as experimental datasets are collected.•Ablation study and sensitivity analysis are conducted to verify the rationality of PMTT architecture.•PMTT has an average improvement of 11.2%, 9.0%, and 9.3% in performance over the optimal baseline model.