| Summary: | 碩士 === 崑山科技大學 === 企業管理研究所 === 93 === This study attempts to evaluate the option price forecasting accuracy of back-propagation neural networks. Our experimental results show that the choices of the number of input nodes, the number of hidden nodes, and the learning rate can affect the forecasting capability of a back-propagation neural network. No matter whether MAPE or MSE is used to rank the performances of the neural networks built, the ranking is the same.
Based on the data used in this experiment, the forecasting performance of a neural network is best when two input nodes, one hidden nodes, and a learning rate of 1 are used. But this does not mean this combination of parameters can be applied in other circumstances. What the experiment demonstrates is that the choice of neural network parameters is an important factor in influencing forecasting performance. Consequently, choosing suitable neural network parameters can be a crucial factor in improving the forecasting capability of a neural network.
|
|---|
