Fuzzy hybrid system for forecasting financial time series

Abstract

We propose a fuzzy hybrid system for forecasting time series, based on the automatic fitting method auto. arima included in the forecast package for R. First, we generate predictions and apply fuzzy clustering to identify patterns and tendencies. Then, using inference criteria on the centers of the clusters we end up with a mean forecast. The system allows the inclusion of expert criteria, i.e., the user can set up restrictions on the clustering based on a priori knowledge of the time series. This approach can be applied to any financial time series meeting the requirements of Seasonal Autoregressive Integrated Moving Average (SARIMA) models. The proposed method is implemented in R. Numerical tests on series of loans, accounts, and saving accounts demonstrate the efficacy of the method.