USING STATISTICAL CONTROL TECHNIQUES FOR IMPROVING EXCHANGE RATE PREDICTIVE MODELS Academic Article in Scopus uri icon

abstract

  • Accurately predicting exchange rates is important for a wide range of stakeholders, from businesses to investors to policymakers, since it can be helpful for making informed decisions and managing risks in an increasingly interconnected global economy. Predicting exchange rates is a challenging task that requires understanding of a wide range of economic, political, and market factors, nevertheless, the prevailing view in the literature is that floating exchange rates are better approximated as random walks processes. Prediction models based on machine and statistical learning using economic fundamentals and time series models have been proposed and demonstrated to improve the random walk model. However, these models usually assume that the used predictor variables and series are in statistical control which is not always true and needs to be accounted for during prediction modeling. This research proposes the addition of statistical control techniques as a method for improving statistical learning of the existing models. The main idea is to use control charts as model selectors in the presence of out-of-control variables. By incorporating control charts into the model, the aim is to identify changes in exchange rates that are beyond the expected range of values. The results obtained will serve as an exploratory analysis to determine whether statistical control approaches are helpful for improving exchange rate predictive models either in the magnitude or the direction of the forecast. By identifying patterns and anomalies in exchange rates, the model looks to help users make more accurate predictions and minimize risks associated with currency exchange. © © American Society for Engineering Management, 2023.

publication date

  • January 1, 2023