Optimized Bayesian tensorized neural network affording task failure prediction in cloud environment

doi:10.1016/j.eswa.2025.127538

Return

Optimized Bayesian tensorized neural network affording task failure prediction in cloud environment

2025-06-06

PRE

Senthil Kumar Avinashi Malleswaran *

Robert, Paulpandi

Padmanaban Kuppan

DOI：10.1016/j.eswa.2025.127538

Original

Original request for help

Save

Abstract

En 中文

A large-scale cloud data centre must have low failure occurrence probability, better service dependability, and availability. Even though, existing large-scale cloud data centers still face higher failure rates owing to a number of issues, like hardware and software malfunctions that result in task and job failures. Several machine learningdependent models have presented for predicting the task or job failure by examining past system message logs. These methods frequently suffer from inadequate prediction accuracy, leading to delayed or inaccurate failure detection. To overcome these limitations, this manuscript proposes an Optimized Bayesian Tensorized Neural Network Affording Task Failure Prediction in Cloud Environment (TFP-BTNN-BBHO). The aim of TFP-BTNNBBHO method is to forecast whether the tasks and jobs are successful or unsuccessful. Initially, the input data is collected from Google cluster traces. Then the input data is pre-processing under Distributed Nonlinear Polynomial Graph Filter (DNPGF). Here, Bayesian Tensorized Neural Network (BTNN) is used to predict the task as failure or success. Finally, Boost Binary Harris Hawks Optimization (BBHO) is proposed to tune the parameters of the BTNN, and improving its prediction accuracy and efficiency. The experimental results show that the proposed method has better precision, recall, and F1-score. The proposed TFP-BTNN-BBHO model attains 99.5% better accuracy for failure and 99.1% for success. It highlights the superior task failure prediction ability of the TFP-BTNN-BBHO model compared with existing methods.

Keywords:

Bayesian tensorized neural network

Boost binary Harris hawks optimization algorithm

Cloud environment

Nonlinear polynomial filtering

Task dataset

Task failure