Development of an AI-Based Model for Predictive Maintenance of Software Systems

Abstract

Software failures have been on the rise recently, leading to operational shortcomings and substantial financial losses. One primary reason is a lack of a diligent maintenance culture. To enhance reliability and minimise operational downtime, this study proposes the development of an AI-based predictive maintenance model for software systems. The study uses real-time and historical system data, such as performance metrics, error logs, and resource usage, in conjunction with supervised machine learning algorithms of Random Forest (RF), Support Vector Machine (SVM), XGBoost, and Long Short-Term Memory (LSTM) networks to forecast software failures before they happen. The researchers used open-source datasets and industrial maintenance logs to train and evaluate the models. Results showed that LSTM and XGBoost performed better than the other models, achieving validation accuracies of up to 90% and demonstrating their strong ability to capture intricate feature interactions and temporal dependencies. By comparing these models for efficacy across comparable scenarios, the study advances the study of software reliability. It emphasises the potential of AI-driven predictive maintenance to help organisations shift from reactive to proactive maintenance strategies, minimising downtime, optimising resource allocation, and lowering costs. Adopting scalable, interpretable models suited to specific data types and system contexts is one of the recommendations.