Trend-Oriented Training for Neural Networks to Forecast Stock Markets

使用趨勢導向的方式來訓練類神經網路以預測股市趨勢

李俊德(Chun-Teh Lee) ； 曾家翔(Jia-Shiang Tzeng)

《Asia Pacific Management Review》 18卷2期 (2013/06) Pp. 181-195

https://doi.org/10.6126/APMR.2013.18.2.04

摘要

本研究使用一種趨勢導向的方式來訓練類神經網路，以增進技術分析對股市趨勢預測的可靠性。根據股價的移動平均線可定義出股市的趨勢值，做為類神經網路預測系統學習訓練時的目標值。我們的測試採用三個基本的技術指標及它們各種不同的組合，這三個基本的技術指標是乖離率（Bias）、威廉指標（%R）、指數平滑異同移動平均線（MACD）。在一段兩年的測試區裡，使用單一技術指標時其趨勢預測準確率可達70%；當多個技術指標被組合在一起使用時，其準確率可增進到74.3%，如果排除中性的平盤趨勢，其準確率更可高達83.1%。我們依據系統預測的股市趨勢與轉折點擬定簡單的交易策略，用來評估投資報酬率。驗證結果顯示，趨勢導向的類神經網路預測系統所獲得的投資報酬遠高於價格導向的參考類神經網路與買入持有策略。證明趨勢導向的訓練方法能提供適當的學習訓練目標，有效的挖掘變動頻繁的股市型態，進而找出其價格趨勢。

關鍵字

移動平均線 ； 類神經網路 ； 股市趨勢 ； 趨勢導向的訓練

並列摘要

A trend-oriented training is developed for neural network training to enhance technical analysis predictions of stock market trends. Predefined trend targets based on a moving average are used for machine learning. Three technical indicators, Bias, %R, and MACD, and their combinations are chosen as input signals to test the system. Accurate prediction rates of about 70% are achievable over a two-year out-of-sample test period when one single input indicator is used. When multiple input indicators are properly combined, the accurate prediction rates can be improved to 74.3% or 83.1% if neutral trends are excluded. A simple trading strategy based on the turning points predicted by the system yields much higher investment returns than a reference neural network system and the buy-and-hold model. The experimental results have demonstrated the effectiveness of this trend-oriented training which gives the system reliable learning targets to capture dynamic price trends.

並列關鍵字

Moving averages ； neural networks ； stock market trends ； trend-oriented training

參考文獻

Armano, G., Marchesi, M., Murru, A. (2005) A hybrid genetic-neural architecture for stock indexes forecasting. Information Sciences, 170, 3-33.

Braun, H., Chandler, J. (1987) Predicting stock market behavior through rule induction: An application of the learning-from-example approach. Decision Sciences, 18, 415-429.

Chang, P.C., Liu, C.H., Lin, J.L., Fan, C.Y., Ng, C. (2009) A neural network with a case based dynamic window for stock trading prediction. Expert Systems with Applications, 36, 6889-6898.

Chen, Y., Mabu, S., Shimada, K., Hirasawa, K. (2009) A genetic network programming with learning approach for enhanced stock trading model. Expert Systems with Applications, 36, 12537-12546.

Freitas, F.D., De Souza, A.F., de Almeida, A.R. (2009) Prediction-based portfolio optimization model using neural networks. Neurocomputing, 72, 2155-2170.