A Dynamic Weighted Distancedbased Fuzzy Time Series Neural Network with Bootstrap Model for Option Price Forecasting

The option price forecasting is still a big challenging problem because the option pricing is determined by many factors. Accordingly, it is difficult to predict option price accurately. To counter this problem, this paper proposes a novel hybrid model to forecast the option price. The proposed model, termed as the dynamic weighted distance-based fuzzy time series neural network with bootstrap model, is composed of a dynamic n-order 2-factor fuzzy time series model, a radial basis function neural network model and a bootstrap method. In the proposed model, the dynamic n-order 2-factor fuzzy time series model can automatic choose the best n-order for searching similar data from historical data and, then, build a training dataset for the radial basis function neural network model to forecast the option price. However, the sample size of option price data is small. Accordingly, this paper uses the bootstrap method to enhance the prediction accuracy of the proposed model. The experiment results show that the proposed model outperforms several existing methods in terms of RMSE, MAE and the testing results of Diebold-Marioano test.