An Interval Technical Indicator for Financial Time Series Forecasting

In this work we compare the performance of some standard technical indicators with an interval technical indicator, the moving interval (MI), for time series forecasting. MI has the advantage of taking into account the variability of data in the range considered and not only the average, like standard indicators do. However, the use of intervals as input variables require the use of regression methods able to handle with non Euclidean structures. The kernel approach is employed to this end. A recently introduced interval kernel is applied together with the moving interval indicator. The conclusion is that this indicator outperforms the forecasting performance of standard indicators.