决策树及神经网络算法在股票分类预测中的应用

发布时间：2018-01-01 22:40

本文关键词：决策树及神经网络算法在股票分类预测中的应用　出处：《杭州电子科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：股票市场作为市场经济的“晴雨表”反映着我国经济的总体状况，在我国经济发展中起着重要的作用。随着股票市场的发展，越来越多的人选择投资股票。为了可以准确的选择优秀的上市公司进行投资，从中获取可观的收益，这就需要对股票市场上不同的上市公司的综合经营绩效进行准确的分析预测。然而股票市场数据量庞大，是一个非常复杂的系统，利用传统的方法很难对它做出准确的预测。数据挖掘技术是一个从海量的杂乱无章的数据中提取出隐含和潜在的对决策有价值的信息和模式的过程，它可以处理股票市场上规模巨大、繁琐、杂乱无章的数据。本文利用数据挖掘技术中的C5.0决策树、BP神经网络和RBF神经网络三种分类算法以上市公司的综合经营绩效等级为分类标准进行分类预测。本文选取2012年A股市场上200个上市公司为样本，其中50个为A股市场上综合绩效最优的股票，50个为综合绩效最差的股票，另外100个为随机选取的综合绩效一般的股票，其中50个为上证股票，50个为深证股票。以股票的综合绩效等级为输出变量，选取七大类14个有代表性的财务指标作为输入变量，运用SPSS Clementine软件，利用C5.0决策树、BP神经网络和RBF神经网络三种分类算法分别建立分类预测模型。在建立模型时，随机选取样本集中80%的数据作为训练样本，用于模型的建立；选取样本集中其余20%的数据作为测试样本，用于模型有效性的检测。模型建立之后，，对三种分类方法的预测准确率进行比较可知， C5.0决策树算法得到的对测试样本集的预测准确率最高，运用C5.0决策树更具有参考意义。由三种分类方法给出的重要变量可知，每股收益增长率对上市公司的综合经营绩效影响最大，现金流动负债比和流动比率对上市公司的综合经营绩效影响也较大。利用三种分类预测模型对上市公司的综合经营绩效进行分析，找出优秀的上市公司财务指标所共有的特征，为投资者在股票的投资决策上提供帮助。
[Abstract]:The stock market as a market economy "barometer" reflects the overall situation of the economy of our country, plays an important role in the economic development of our country. With the development of the stock market, more and more people choose to invest in stocks. In order to accurately select the best listed companies to invest in, get considerable profit from this. You need on the stock market on different listed companies comprehensive performance analysis accurately predict the stock market. However, the huge amount of data, is a very complex system using traditional methods, it is very difficult to make accurate predictions. Data mining technology is an out of order from the mass data extraction process the implicit and potentially valuable information for decision-making and mode, it can be a huge scale, the stock market is complicated, out of order data. By using data mining technology. The three classification algorithms of C5.0 decision tree, BP neural network and RBF neural network are classified and forecasted according to the classification standard of listed companies' comprehensive management performance.
This paper selects 200 listed companies in 2012 A stock market as samples, of which 50 are A stock market performance optimal stock, 50 is the worst performance of the stock, the other 100 were randomly selected for the comprehensive performance of common shares, 50 of which the Shanghai stock, 50 for the Shenzhen stock. In order to grade comprehensive performance stock as the output variables, selected seven categories 14 representative financial index as input variables, the use of SPSS Clementine software, using C5.0 decision tree classification, prediction models are established BP neural network and RBF neural network three classification algorithms. In the models, randomly selected sample of 80% data as the training samples, used for model establishment; the remaining 20% of the sample concentration data as test samples for the detection of the effectiveness of the model. After the model, the prediction of three classification accuracy ratio Is the prediction of the test set of C5.0 decision tree algorithm is the highest accuracy, using C5.0 decision tree has the reference significance. It presents three important variables by the classification method, the greatest impact on the growth rate of earnings per share of listed companies comprehensive operating performance, influence of cash flow debt ratio and liquidity ratio of the performance of listed companies is also larger. On the listed company's comprehensive performance is analyzed by using the three classification prediction model, find out the financial index of listed company's outstanding common characteristics, help provide investors in the investment decision-making stock.

【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.51;TP183

【参考文献】