【作者】 韩莉；

【导师】 王福林；

【作者基本信息】 东北农业大学 ， 管理科学与工程， 2016， 硕士

【摘要】 股票市场经过数载发展,在市场经济中占据了越来越重要的地位。股票市场的建立和发展,不仅受国家经济的影响,也为国家的经济建设时时刻刻在做着贡献。然而,股票投资市场并不十分稳定和平稳,成交量和价格方面总有着意料外的波动。股票市场是股票投资者进行交易的平台,无形中在投资者和筹资者之间搭建了利润提升的桥梁。在股票市场上,筹资者公开募股,发行股票,为长期的资金来源提供了保障;与此同时,投资者通过购买公开募股的股票,相当于与公司共发展,上市公司的优劣直接会影响投资者的收益。由于投资者们的心理状态以及投资偏好都不尽相同,故也会选择不同的投资组合,也会承担不同的投资风险。然而,这样的投资也不是一直稳赚不赔的,股票市场波动性很大,投机成分强,股票市场缺乏效率,稳定性差,这些都会危及到股票本身的进一步发展。股票市场的效率体现在上市公司能够合理分配资金并将资金的利用率提升到最大,从而争取更多利润的能力。可是,大量的实证证明,股票市场并不是十分有效的。然而,股价的走势也有一定的规律性可言,这体现在这种走势可以通过非线性函数进行描述,那么也就是可以预测的。影响股价的因素多种多样,对股票所起的作用也复杂多变,为了更加准确的进行预测,将人工神经网络引入到了金融预测领域。原则上,对于连续函数,神经网络能在一定的精度范围内实现良好的训练。人工神经网络可以解决黑箱问题,它回避了数据变化的内在原因,更加科学地通过特定的学习样本进行机器训练,建立一种模型来描述输出与输入变量之间的联系。因此,研究基于神经网络的股票预测问题,不仅具有理论意义,也具有重要的现实意义和参考价值。本文对现有的股票预测方法、BP神经网络及其存在的问题、LM-BP神经网络算法、LM-BP神经网络对股价预测等问题进行了系统研究。在研究过程中,取得的成果主要有:(1)对股票市场特点和股票预测方法进行分析,指出了这种方法的优缺点。(2)针对股票价格预测数据量大,应用标准BP神经网络运算速度慢的问题,推导给出了LM-BP神经网络算法,并设计开发了LM-BP神经网络计算程序。(3)应用LM-BP神经网络预测了美国纳斯达克证券交易所挂牌上市的智联招聘股票的开盘价、最高价、最低价、收盘价走势。预测结果表明,预测精度较高,开盘价格平均相对误差为0.88%,最高价平均相对误差为1.25%,最低价平均相对误差为1.26%,收盘价平均相对误差为1.4%。(4)在预测的基础上,计算给出了移动平均线(MA)、乖离率(BIAS)、相对强弱指标(RSI)、随机指标(KDJ)、人气指标(BOV)、威廉指标(W&R),并画出了移动平均线(MA)、乖离率(BIAS)、相对强弱指标(RSI)、随机指标(KDJ)、人气指标(BOV)、威廉指标(W&R)曲线,为投资者决策提供参考。最后,本文对股票市场预测问题进行了展望。更多还原

【Abstract】 The stock market after several years of development, in a market economy occupy an increasingly important position. The establishment and development of the stock market, not only affected the national economy, but also for the country’s economic construction always doing contribution. However, the equity market is not very stable and smooth, with a total trading volume and price volatility of the unexpected. Stock market investors is the stock trading platform,virtually between investors and fund-raisers to build a bridge to enhance profits. In the stock market, fund-raisers public offering, issuing shares, long-term sources of funding has provided a guarantee; At the same time, investors by purchasing IPO shares, equivalent to the total development of the company, listed companies will directly affect It returns to investors. Due to the psychological state of investors and investment preferences are different, they will choose different portfolios, will assume different investment risk. However, such investment is not has not lose, stock market volatility is large, strong speculative stock market inefficiency, poor stability which will jeopardize the further development of the stock itself. Efficiency is reflected in the stock market listed companies to rational allocation of funds and the utilization of the funds raised to the maximum, so the ability to strive for more profits. However, a large number of empirical proof, the stock market is not very effective. However, stock prices have a certain regularity at all,which is reflected in this trend can be described by a nonlinear function, it is predictable. A variety of factors that affect the stock price, the stock is also the role of the complex, in order to more accurately predict the artificial neural network is introduced to the field of financial forecasting. In principle, for a continuous function, the neural network can be trained to achieve good accuracy in a certain range. Artificial neural networks can solve the problem of the black box, it sidesteps the underlying causes of data changes, more scientifically trained through a specific machine learning samples, establish a model to describe the connection between the input and output variables.Therefore, the study Stock prediction based on neural networks, not only of theoretical significance, but also has important practical significance and reference value.In this paper, the existing stock prediction method, BP neural network and its problems,LM-BP neural network algorithm, LM-BP neural network to predict stock price and other issues has been systematically studied. During the study, the results achieved are:(1) the characteristics of the stock market and stock forecasting methods to analyze the advantages and disadvantages of this approach.(2) to predict the amount of data, using standard BP neural network computing problems of slow for the stock price, derived gives LM-BP neural network algorithm, and design and development of LM-BP neural network calculation program.(3) Application of LM-BP neural network to predict the NASDAQ Stock Exchange-listed shares Zhaopin open, high, low, closing price trend. The prediction results show higher precision,the opening price of the average relative error is 0.88%, the highest price average relative error is1.25%, the lowest average relative error is 1.26%, the closing price of the average relative error of1.4%.(4) the forecast on the basis of the calculation is given Moving Average(MA), a deviation rate(BIAS), Relative Strength Index(RSI), stochastics(KDJ), sentiment indicator(BOV), William indicators(W & R), and draw a moving average(MA), a deviation rate(BIAS), relative strength index(RSI), stochastics(KDJ), sentiment indicator(BOV), William indicators(W & R) curve for investor decision-making for reference.Finally, the stock market prediction in the future.更多还原