Utilization of various time series models forecasting gold prices in Thailand

Abstract

This work develops an interactive dashboard integrating various time series models to forecast gold prices in Thailand, enabling investors to make better decisions more efficiently and effectively manage their investments. The study used monthly data from 2009 to 2021, separated the data series into in- and out-samples, and found that the gold price dataset did not have an autoregressive conditional heteroskedasticity (ARCH) effect. The process consumed the autoregressive integrated moving average (ARIMA) best model without passing through the generalized autoregressive conditionally heteroskedastic (GARCH)/ARIMA best model. The highest-performance model for forecasting was the ARIMA (1,1,1) model. This research extends the implementation scope of previous research on gold price forecasting with developer fulfillment in Thailand by developing a business intelligence dashboard for users to utilize the predictions. The dashboard is interactive, allowing users to filter the data and predictions based on their needs. Integrating various time series models for forecasting gold prices in Thailand on a single dashboard will enable investors to make better decisions and manage their investments efficiently and effectively. The authors are also developing an automatic utilization script to further improve the dashboard usability.