| Summary: | 碩士 === 雲林科技大學 === 財務金融系碩士班 === 97 === This study utilizes an artificial neural network (ANN) to integrate eight popular technical indicators (KD, RST, MACD, AR, W%R, MTM, TAPI, PSY) for emitting trading signals in the Taiwan stock market. Moreover, this research also employs a network pruning method to simplify the network connection and classifies the resulting model into simple trading rules. The purposes of this study mainly are the evaluating the ANN integrated technical trading system and the efficiency of the network pruning technique. The empirical research deals with the Taiwan stock index from the period January 18, 2000 to April 3, 2008 and with five moving window in which 300 daily closing prices are drawn as trading set and forward 100 closing prices are serves as testing period. The major results are indicated in the following statistics: the average returns of the benchmark are -25.09%, 6.66%, 13.51%, 5.45%, and -5.65% in the five windows, respectively; contrastingly, the ANN based trading system being 41.70%, 50.23%, 59.24%, 21.25%, 54.05%; the punned network being 72.115, 34.67%, 41.91%, 8.06%, and 120.03%; the average returns of eight technical indicators are 14.654%, 18%, 2.41%, 9.22%, and 4.03%.
The empirical evidence shows that the ANN based technical trading system is not only capable of beating the Taiwan stock index but also significantly outperforms the individual technical indicator. Further, the pruned network can in average exceeds the original network in terms of the average returns of five windows: 55.36% vs. 45.29%.
|
|---|
