| Summary: | 碩士 === 長庚大學 === 企業管理研究所 === 95 === The Multilayer Back Propagation Neural Network (MLBPN) applied to establish a forecast model, used to predict the variation in sector stock index of Taiwan. According to interview the fund manager, there are 10 Technical Indexes usually used by the fund manager and have higher relation factors, and which are applied to be the input variables for the neural predictive model. The output variables are the variations of 2、5、10、15 and 22 days stock indexes. The data pairs of 230 and 1095 days are used to train the neural models. The data pair of 230 days also used to sliding window training. The following 175 days Technical Indexes and the variations of 2、5、10、15 and 22 days stock indexes are used as testing data pairs.
The simulation results reveal the prediction of 2 days variations of stock indexes is best accurate with least convergent error, and 5, 10, 15 and 22days prediction is second, third, forth, fifth respectively. It reveals that more days prediction will get more predictive error. And 1,095 days training have the best accuracy of 2 days prediction. It means more information for training the neural model will obtain better predictive accuracy.
|
|---|
