| Summary: | 碩士 === 雲林科技大學 === 財務金融系碩士班 === 98 === For financial investors, a challenging task is determining market timing—when to buy and sell a stock. Due to the complexity of stock market data, the prediction of stock price can be a very difficult task. In this study, two learning paradigms of neural networks, supervised versus unsupervised, are compared using their representative types of Backpropagation network (BPN) and Kohonen self-organizing feature map (Kohonen SOM). Further, this paper also proposes a hybrid model by integrating Kohonen SOM with BPN to predict the TWSE TAIEX index. The empirical results illustrate the standard derivations of Kohonen SOM, BPN, and the hybrid model respectively account for 16.51%, 17.39% and 16.19%. The evidence demonstrates the proposed model is more robust than Kohonen SOM and BPN.
|
|---|
