Visualization and Forecasting of time series data is difficult when the data are very numerous, with complex structures as, for example, in the presence of high volatility and structural changes. This is the case of high frequency data or, in general, of financial data, where we cannot clearly visualize the single data and where the necessity of an aggregation arises. In this paper we deal with the specific problem of forecasting beanplot time series. We propose an approach based firstly on a parameterization of the beanplot time series and successively on the chosen best forecasting method with respect to our data. In particular we experiment with a strategy to use combination forecast methods in order to improve the forecasting performance.
- Time Series
Available at: http://works.bepress.com/carlo_drago/101/