基於深度學習與使用技術分析跟頻率分解的股票價格預測

Stock price prediction based on deep learning and using technical analysis and frequency decomposition

張家齊(Chia Chi Chang)

指導教授 ： 劉冠顯

共同指導教授 ： 盧永豐(Yung-Feng Lu)

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2022年)

摘要

股票預測一直都是金融界感興趣的話題，合理準確的預測能提高財務收益並且對抗市場風險。但由於股票市場有著非線性、非平穩以及高波動性，因此準確的預測一直是一項挑戰。投資人觀察市場變化，股市走勢以及各個消息面，不斷的去調整投資策略進行股票買賣，從而獲得財務回報。
但是金融市場是非線性、非平穩與高波動性的數據。為了能夠精準預測股票價格，本研究提出了一種基於1DCNN-Attention的深度神經網路模型來捕捉深度序列關係，並且預測第二天的股價。其金融數據是使用專家多年經驗的技術分析指標與完整經驗分解的方法所結合的混合數據，來試圖捕捉金融中的周期趨勢與規律。
實驗結果表明，這種混合特徵提取與基於1DCNN-Attention的架構獲得了不錯的性能。此架構也表明了比起傳統的時間序列方法如RNN、LSTM等訓練速度更快。因此所提出的模型可以減少更多的訓練時間與機器成本，優於其他最先進的方法。

關鍵字

一維卷積神經網路 ； 注意力機制 ； 完全經驗模態分解 ； 頻率分解 ； 股票預測 ； 技術分析

並列摘要

Stock forecasts have always been a topic of interest in the financial community. Reasonable and accurate forecasts can improve financial returns and counter market risks.
However, due to the non-linear, non-stationary and high volatility of the stock market, accurate forecasting has always been a challenge
Financial market data is non-linear, non-steady and highly volatile. In order to obtain higher stock forecast accuracy, we propose an 1DCNN-Attention based deep neural network model to capture the deep sequence relationship analysis and predict the stock price of the next day. We also use a mixed feature extraction by combining expert technical analysis indicators and data decomposition with complete empirical mode to capture the trend law in financial data.
The experimental results show that this hybrid feature extraction and 1DCNN-Attention based architecture achieve good performance. It also shows that this architecture can achieve high training speed, and its performance is superior to traditional time series methods such as RNN and LSTM. Hence, the proposed model can reduce more training time and machine cost than other state-of-the-art methods.

並列關鍵字

1DCNN ； Attention ； CEEMD ； Frequency Decomposition ； Stock Forecast ； Technical Analysis

參考文獻

[1] Eugene F. FAMA, "The Behavior of Stock-Market Prices," The Journal of Business, vol. 38, no. 1, pp. 34-105, 1965.

[2] Yaser S. Abu-Mostafa and Amir F. Atiya , "Introduction to financial forecasting," Applied Intelligence, vol. 6, pp. 215-213, 1996.

[3] Y. LeCun, Y. Bengio and G. Hinton, "Deep learning," Nature, no. 521, pp. 436-444, 2015.

[4] Jeffrey L. Elman, "Finding Structure in Time," Cognitive science, vol. 14, no. 2, pp. 179-211, 1990.

[5] Andrés Vidal and Werner Kristjanpoller, "Expert Systems with Applications," vol. 157, 2020.