A Data-driven Inspection of the Silicon Price Market

Abstract

This thesis explores the evolving demand dynamics of metallurgical grade silicon in three different regions, the US, European Union, and China. It leverages statistical and machine learning methodologies to predict future trends. Silicon, a critical element with widespread applications, is utilized differently based on its purity—from simple building materials to sophisticated electronic components. By employing Autoregressive Integrated Moving Average (ARIMA) models, Generalized AutoRegressive Conditional Heteroskedasticity (GARCH), Vector Autoregression (VAR), and advanced AutoML techniques, this thesis aims to provide a insight into effective methods for predicting the volatile silicon market. We find that due to the sparse historical price reporting and unforseen factors around Covid, successfully predicting future trends exhibits some challenges.