基于优化BP神经网络和粒计算的股指预测研究

发布时间：2018-05-07 21:46

本文选题：BP神经网络 + 遗传算法　；参考：《南京大学》2012年硕士论文

【摘要】：根据有效市场的假说短期股价是随机游走的,不具有可预测性,任何试图预测的技术手段都只能是徒劳无益的。然而大量证据表明市场并非有效。第一,投资者理性的假说就遭到质疑,首先有效市场假说认为所有的投资者都会对市场公开信息以差不多的速度,做出相同的反应,在现实市场中并非如此。第二,大量的研究表明股价波动的趋势是存在的,且与重大事件和经济指标有相关性。因此股价具有一定的可预测性,由于影响股价的因素太多,且作用机理很复杂,传统的方法很难奏效。神经网络以其优秀的模拟特性在股价预测方面占有重要的一席之地。本文采用应用最广泛的BP神经网络来做预测,根据预测结果发现问题,针对其缺陷,采用遗传算法进行优化,并对照两次预测的结果。针对神经网络只能进行单点预测,没有办法进行未来一段时间趋势和范围预测本文引入当前的一个备受关注的热门研究领域——粒计算,从而解决了上面的问题。本文得出以下几点结论： (1)良好的预测结果表明中国股票的可预测性,在一定程度上证明了中国的股票市场还不是有效市场 (2)运用遗传算法优化BP神经网络的初始权值和阈值,得出了更好的结果,表明这种方法的可行性。 (3)将粒计算和优化的BP神经网络相结合预测的结果总体上令人满意,表明粒计算在金融预测上是可行的,为理论和实践都能提供有益指导。
[Abstract]:Based on the efficient market hypothesis that short-term stock prices are random and unpredictability, any attempt to predict the technical means can only be futile. Yet there is plenty of evidence that markets are not efficient. First, the hypothesis of investor rationality is questioned. First of all, the efficient market hypothesis assumes that all investors will react to the market disclosure information at the same speed, which is not true in the real market. Second, a large number of studies show that the trend of stock price volatility exists, and is related to major events and economic indicators. Therefore, the stock price has a certain degree of predictability, because there are too many factors affecting the stock price, and the mechanism is very complex, the traditional method is difficult to work. Neural network plays an important role in stock price prediction for its excellent simulation characteristics. In this paper, the most widely used BP neural network is used to make prediction. According to the prediction result, the problem is found, and the genetic algorithm is used to optimize the problem according to its defects, and the results of twice prediction are compared. The neural network can only carry out single point prediction, but there is no way to predict the trend and range for some time in the future. In this paper, grain computing, a hot research field that has been paid close attention to, is introduced in this paper, thus solving the above problem. This paper draws the following conclusions: Good forecast results show the predictability of Chinese stocks, and to some extent prove that China's stock market is not an efficient market. 2) genetic algorithm is used to optimize the initial weight and threshold of BP neural network, and a better result is obtained, which shows the feasibility of this method. 3) the result of combining particle computing with optimized BP neural network is satisfactory, which indicates that particle calculation is feasible in financial forecasting and can provide useful guidance for both theory and practice.
【学位授予单位】：南京大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F830.91;F224

【相似文献】