Development of a simulation result management and prediction system using machine learning techniques

Abstract

Simulations are widely used in various fields of computationalscience and engineering. As IT technology advances, the complexity andaccuracy requirements of the simulations are increasingly rising up,accordingly escalating their execution cost as well. Nevertheless, it appears thatthe community has not yet paid much attention to the reuse of previouslyobtained simulation results to improve the performance of the execution oflater requested simulations. In this regard, we propose a novel simulationservice system that can utilise the results of previously executed simulationsand thus improve the performance of later simulations. The proposed systemcan not only convert completed simulation results into a standard form andthen store them into a NoSQL database for efficient retrieval, but also predictthe result of a requested simulation using machine learning techniques withoutactual simulations. We demonstrate that the proposed system achieved verylow error prediction rates only up to 7.4% from 0.9%.