由於人工智慧的高度發展,因此實務界逐漸採取相關人工智慧處理複雜的巨量數據(big data)。本研究運用了也在人工智慧範疇類的類神經網路,並結合了技術分析指標數據做為類神經網路的預測變數,而類神經網路擁有優秀的容錯能力,即使有雜訊資料也能產生較為精確的預測結果,已有許多研究及文獻採用類神經網路為研究工具,進而提升預測能力。本文首先比較不同隱含層與不同神經元數組合之預測績效,接著比較加入GARCH波動度作為類神經網路預測變數後的績效,並採用摩根台灣指數作為預測標的,使用其2014年至2018年之歷史股價資料探討預測績效。最後結果顯示,類神經網路設定為2×2在預測5天期時,已具有良好的預測績效。本研究可供使用類神經網路預測價格、波動等等研究為參考依據,特別是以技術分析指標作為變數進行價格預測之研究。
Following the development of artificial intelligence, practitioners are gradually using artificial intelligence to handle complex huge amounts of data (big data). This study uses a neural network that is also a kind of artificial intelligence technology and combines with technical analysis data as the predictive variables of artificial neural network. First, we compare the predicted performance of different hidden layer and different neuron number combinations, and then compare the performance of GARCH volatility as the predictive variable of artificial neural network. We use the historical share price data of MSCI Taiwan Index for the period of 2014-2018 and view the predicted performance through historical share price data. The artificial neural network with excellent fault tolerance, even if there is noise data also can produce more accurate prediction results. Many previous studies and literature used artificial neural networks as a research tool to improve predictive performance. This study can be used as a reference basis for the prediction of the price, volatility, etc. by using artificial neural networks. In particularly, the technical analysis of data as a variable for price forecasting research.