Design and Development of a Predictive Digital Twin for Proactive Cloud Resource Management: Application to EGovernment Services

Abstract

Digital Twin technology enables the creation of virtual representations of physical systems for monitoring and management purposes. When combined with Artificial Intelligence, Digital Twins can provide predictive capabilities that support proactive decision-making. This thesis proposes a Predictive Digital Twin framework for proactive cloud resource management in e-government services. The framework integrates a Digital Twin platform, predictive analytics, and explainable artificial intelligence to predict future scaling decisions based on historical telemetry data. Several machine learning and deep learning models, including Random Forest, XGBoost, BiLSTM, and Transformer, were evaluated using the Alibaba Cluster Trace 2018 dataset. The experimental results demonstrate the feasibility of predicting scaling actions (scale_up, maintain, and scale_down) from cloud resource metrics. The proposed framework extends traditional Digital Twin monitoring by incorporating predictive intelligence and decisionsupport capabilities for cloud administrators