Forecasting volatility with simple linear regression and ordered weighted average operators

Abstract

Estimating and forecasting volatility is an important issue for financial decision-makers. Therefore, it is important to build models that adapt to the current characteristics of the time series. The ordered weighted average (OWA) has some extensions that provide interesting ways to adapt to these characteristics. This work proposes a new application that uses the simple linear regression (LR) and OWA operators in the same formulation. We use the heavy ordered weighted average (HOWA), the prioritized ordered weighted average (PrOWA), the probabilistic ordered weighted average (POWA) and their combinations with induced cooperators. The main idea in linear regression with OWA operator is to obtain an estimate and forecast that can be adaptable to situations of uncertainty and information known to the decision maker. The work analyzes the applicability of this new approach in a problem regarding exchange rate volatility forecasting, where the operators that we can use in high or low seasons are located and thus generate ranges.