Time Series Forecasting using Gaussian Mixture Density

摘要

We present a new appraoch for financial time series forecasting using a mixture of Guassian density mdoels Because of the probabilistic formulation, it has many useful properties not found in other traditional approaches. For example, a meaningful refjection criterion can be implemented so that the forecasting system does not need to make a recommendation when it is uncertain about the future. Experimental results will be presented to demonstrate the usefulness of this appraoch.