基于深度学习方法的股票市场预测

STOCK MARKET FORCAST BASED ON DEEP LEARNING METHOD

戴睿

11749005

硕士

金融

李昕

2019-05-27

2019-06-28

哈尔滨工业大学

深圳

众所周知，股票市场是一个变幻莫测的市场，其上涨和下跌受到许多因素的影响。随着中国经济的发展，越来越多的人将目光投向股票市场，以期分享祖国发展的红利。但是股票市场的涨跌并不直接受到经济大环境影响，而是会受到很多影响因素的制约，不仅公司本身的业绩和发展前景对股价会有影响，投资者们的信心以及大型投资公司的态度同样影响很大。找寻股价涨跌的秘密成为不少研究人员追求的目标。在此之前，人们运用的方法一般是传统的统计学方法，比如直观的K线图，MA，MACD趋势线等。但是这些传统方法虽然直观但是并不能准确有效的反映市场情况。随着机器学习技术的兴起，量化金融迎来了新的发展机遇，预测的准确率得到大幅的提升。大数据时代来临，数据量得到极大拓展的情况下，深度学习技术也得到了相应的发展，因此本文使用深度学习技术对股票进行预测。使用的主要方法是基于改进的长短期记忆网络（Long-Short Term Memory，简称LSTM）模型，并使用Adam方法进行优化，对Amazon的股票数据进行相关预测，计算出预测的准确率并画出图形，观察其计算的准确率。计算发现，可以获得非常良好的效果，能够为投资者们提供投资帮助。最后对这种方法进行分析并提出了可能的改进方案，为后人的研究提供指导方向。

As we all know, the stock market is a changeable market, its rise and fall are affected by many factors. With the development of China's economy, more and more people are turning their eyes to the stock market in order to share the dividends of the development of the motherland. However, the rise and fall of the stock market is not directly affected by the economic environment, but will be constrained by many factors, not only the company's own performance and development prospects will have an impact on the stock price, but also the confidence of investors and the attitude of large investment companies. Searching for the secret of stock price fluctuation has become the goal of many researchers. Previously, people used traditional statistical methods, such as intuitive K-line graph, MA, MACD trend line and so on. But these traditional methods are intuitive, but they can not accurately and effectively reflect the market situation. With the rise of machine learning technology, quantitative finance has ushered in new opportunities for development, and the accuracy of prediction has been greatly improved. With the advent of the era of big data and the great expansion of data volume, deep learning technology has also been developed accordingly. Therefore, this paper uses deep learning technology to predict stocks. The main method is based on the improved Long-Short Term Memory (LSTM) model of cyclic neural network, and the Adam method is used to optimize it. The stock data of Amazon are forecasted, and the accuracy of prediction is calculated and graphed, and the accuracy of calculation is observed. The calculation shows that it can achieve very good results and provide investors with investment assistance. Finally, this method is analyzed and possible improvement schemes are proposed, which can provide guidance for future research.

股票走势预测 深度强化学习 长短期记忆网络

stock price prediction deep reinforcement learning Long-Short Term Memory

中文

联合培养

南方科技大学

戴睿. 基于深度学习方法的股票市场预测[D]. 深圳. 哈尔滨工业大学,2019.