資料探勘及計算智慧技術應用 於股價交易訊號之預測研究

Apply Data-mining and Computational Intelligence for Stock Trading Signals Predication

樊晉源(Chin-Yuan Fan)

指導教授 ： 張百棧

元智大學/工程學院/工業工程與管理學系/博士(2009年)

摘要

不同於傳統的金融時間序列預測的工具，計算智慧技術近年來逐漸取代前者並應用於金融時間序列預測，與傳統的金融時間序列預測的工具相比，計算智慧技術不但可以快速找到最佳解，其解答之品質也更為可靠。本研究嘗試結合資料探勘工具和計算智慧技術處理金融時間序列數據之預測。本研究首先運用小波理論及動態時間規化對財務資料進行前處理，之後進一步套用資料探勘之分群手法及不同之計算智慧方法解決不同之股票預測類型之問題，更進一步運用PLR（線段切割法）和模糊決策樹檢測不同的股票市場，進而探討其最佳交易時間。其最佳交易時間之檢測和股票市場決策買賣訊號之預測藉由叢集式神經網絡的訓練而判斷。實驗結果證實，藉由本研究財務系統之測試，不論上升趨勢，穩定和下降趨勢之股票，均有令人讚賞之績效。未來更可嘗試將此一系統投入市場，進行實際之判斷與應用。

關鍵字

資料探勘 ； 財務資料預測 ； 計算智慧工具 ； 資料分群 ； 交易訊號判斷 ； 股價預測

並列摘要

Recently, researchers have applied computational intelligence techniques in financial time series forecasting. Unlike traditional financial time series forecasting tools, Computational intelligence techniques can find optimal solutions reliably and quickly. This research combines data mining tools and computational intelligence techniques by using a data mining tools pre-process for financial time series data. This process includes a wavelet pre-process and a dynamic time warping stage. Similar data patterns can be retrieved from historical data to predict future stock prices with these patterns. This paper utilizes different computational intelligence approaches by applying PLR and Fuzzy decision tree to decompose historical data into different segments. As a result, this research detects and inputs temporary turning points (trough or peak) of the historical stock data to the back propagation neural network for supervised training of the model. The current study tests the proposed financial system on different types of stocks, i.e., up-trend, steady, and downtrend. The experimental results show that the proposed system generates significant amounts of profit on stocks with different variations. Therefore, the proposed system is very effective and precise for predicting future trading points of a specific stock.

並列關鍵字

Data Mining tools ； Financial time series forecasting ； Computational Intelligence ； Data Clustering ； Trading Signal Detection ； Stock Prices Prediction

參考文獻

[1]. Abonyi, J., Nemeth, S., Vincze, C., Arva, P. (2003). Process analysis and product quality estimation by Self-Organizing Maps with an application to polyethylene production. Computers in Industry, 52(3), pp.221-234.

[2]. Aiken, M., Bsat, M.(1999). Forecasting Market Trends with Neural Networks. Information Systems Management , 16 (4), pp.42-48.

[3]. Back, B., Sere, K., Vanharanta, H. (1998). Managing complexity in large data bases using self-organizing maps. Accounting, Management and Information Technologies, 8(4), pp.191-210.

[4]. Bao, D.P., Yang, Z.H. (2008). Intelligent stock trading system by trading point confirming and probabilistic reasoning. Expert Systems with Applications, 34, pp.620-627.

[7]. Dempster, M. H. A., Jones, C. M. (2001).A real-time adaptive trading system using genetic programming. Quantitative Finance ,1, pp.397-413.