Artificial Neural Networks (ANNs) are flexible computing frameworks and universal approximators that can be applied to a wide range of time series forecasting problems with a high degree of accuracy. However, despite of all advantages cited for artificial neural networks, they have data limitation and need to the large amount of historical data in order to yield accurate results. Therefore, their performance in incomplete data situations is not satisfactory. Although, no definite rule exists for the requirement of the sample size for a given problem, the amount of data for the network training depends on the network structure, the training method, and the complexity of the particular problem or the amount of noise in the data on hand.
However, collecting the necessary data is very expensive and time-consuming. Therefore, due to rapid changes in real situations and financial and economic systems, especially, forecasting in these environments is needed for methods that are also efficient with less available data. Since in fuzzy forecasting models, we use fuzzy numbers instead of crisp values, this method requires fewer observations and it is suitable under incomplete data conditions. However, their performance is not always satisfactory, especially, when the training data set includes a significant difference or outlying case.
Using hybrid models or combining several models has become a common practice to improve the forecasting accuracy and the literature on this topic has expanded dramatically. In this paper, fuzzy regression models are applied to construct a new hybrid model of ANN in order to yield more accurate model than traditional neural networks; especially, for cases where inadequate historical data is available. In our proposed model, fuzzy numbers are used as parameters values of artificial neural networks (weights and biases), instead of using crisp values. In order to show the appropriateness and effectiveness of the proposed model for time series forecasting, the proposed model has been applied to gold price forecasting problem and their performance has been compared with their components. Empirical results indicate that the proposed model is an effective method in order to improve forecasting accuracy. Therefore, it can be applied as an appropriate alternative model for forecasting tasks, especially when higher forecasting accuracy is needed.