An intelligent pattern recognition model for supporting investment decisions in stock market

详细信息    查看全文

• 作者：Tai-liang Chen^a ; ^{97007@mail.wzu.edu.tw" class="auth_mail" title="E-mail the corresponding author} ; ^{too.cool.chen@gmail.com" class="auth_mail" title="E-mail the corresponding author} ; Feng-yu Chen^b
• 关键词：Stock market forecasting ; Perceptually important point (PIP) identification matching ; Template matching technique ; Chart pattern ; Technical indicator
• 刊名：Information Sciences
• 出版年：2016
• 出版时间：10 June 2016
• 年：2016
• 卷：346-347
• 期：Complete
• 页码：261-274
• 全文大小：1693 K

文摘

For many years, how to make stock market predictions has been a prevalent research topic. To carry out accurate forecasting, stock analysts and academic researchers have tried various analysis techniques, algorithms, and models. For example, "technical analysis” is a popular approach used by common stock investors to analyze market trend, and Artificial Intelligence (AI) algorithms such as genetic algorithms (GAs), neural network (NN), and fuzzy time-series (FTS), were proposed by researchers to forecast the future stock index. Although the daily forecasts are very useful for professional investors who implement intraday trading, we argue that forecasting a bullish turning point is a more interesting issue than the future stock index for common investor because an accurate forecast will bring a huge amount of stock return. Therefore, this paper proposes an intelligent pattern recognition model, based on two new stock pattern recognition methods, “PIP bull-flag pattern matching” and the “floating-weighted bull-flag template,” to recognize a bull-flag stock pattern. The bull-flag pattern is a stock's turning point with proper timing, which can enable a stock investor to profit. To promote recognition accuracy, the proposed model employs chart patterns and technical indicators, simultaneously, as pattern recognition factors. In the model verification, we evaluate the proposed model with stock returns by forecasting two stock databases (TAIEX and NASDAQ), and comparing the returns with other advanced algorithms. The experimental results indicate that the proposed model outperforms the published algorithms, such as rough set theory (RST), genetic algorithms (GAs) and their hybrid model, and gives a high-level of profitability. Additionally, the trading strategies, provided by the proposed model, also help investors to make beneficial investment decisions in the stock market.