Search

Improving forecasting especially time series forecasting accuracy is an important yet often difficult task facing forecasters. Fuzzy autoregressive integrated moving average (FARIMA) models are the fuzzy improved version of the autoregressive integrated moving average (ARIMA) models, proposed in order to overcome limitations of the traditional ARIMA models; especially data limitation, and yield more accurate results. However, the forecasted interval of the FARIMA models may be very wide in some specific Circumstances. For instance, when data has high volatility or includes a significant difference or outliers. In this paper, a new hybrid model of FARIMA models is proposed by combining with probabilistic neural classifiers, called FARIMAH, in order to yield a more general and more accurate model than FARIMA models for financial forecasting in incomplete data situations. The main idea of the proposed model is based on this fact that the distribution of the actual values in the forecasted interval by FARIMA is not uniform. Thus, by detecting the spaces with more probability for actual values using the probabilistic classifier, narrower interval than traditional FARIMA models can be obtained. Empirical results of exchange rate markets forecasting indicate that the proposed model exhibit effectively improved forecasting accuracy, so it can be used as an alternative model to exchange rate forecasting, especially when the scant data made available over a short span of time.