SCALABLE TIME SERIES SIMULATION USING PROXY-BASED MODELING TECHNIQUES

Abstract

The fast development of large-scale sequential data obtained from financial systems, IoT devices, cloud platforms, and industrial applications, has led to an increasing demand for scalable and computationally efficient time series simulation methodologies. Existing simulation approaches are generally limited by their scalability, memory consumption and computing cost when working on large-scale data. In this paper we present a scalable time series simulation framework by applying proxy-based modelling techniques for improving the computing efficiency while preserving the forecast accuracy. The architecture we offer exploits distributed computing technologies, machine learning based proxy models and parallel processing techniques for effective simulation of high-volume time series data. The experimental evaluation was performed on benchmark and generated datasets in a distributed cloud-based setting. The results showed considerable increases in execution speed, scalability efficiency, resource utilisation and prediction accuracy compared with conventional simulation methods. Statistical analysis employing metrics such as Mean Squared Error (MSE), Root Mean Square Error (RMSE) and correlation analysis proved the dependability and efficacy of the suggested framework. In summary, the study demonstrates that proxy-based modelling techniques can be a viable approach to support real-time and large-scale time series simulation applications in modern data-intensive systems.