基于数据挖掘技术的股票市场分析与预测

发布时间：2018-03-12 17:47

本文选题：模糊时间序列　切入点：K均值算法　出处：《吉林财经大学》2017年硕士论文　论文类型：学位论文

【摘要】：股票市场作为证券行业至关重要的组成部分,备受投资者的关注。寻求有效的股票分析方法,降低投资者的风险,具有重大实践意义和理论价值。然而,由于股票市场受股票内在价值、市场因素、政治因素以及宏观经济运行状况等诸多因素影响,各因素间没有确定规则,且股票市场每天产生大量数据,这些均给股票市场研究带来一定难度。数据挖掘结合数据库、统计学和人工智能等多门学科,它能够从大量的原始数据中挖掘出隐含的有价值的信息。股票市场的特征决定了应用数据挖掘方法对股市分析和预测,具有较强的可行性和现实性。本文通过收集整理股票指标及上市公司财务数据,利用数据挖掘中的分类和聚类方法,针对不同问题提出一系列改进的算法,对我国股市进行分析与预测,主要内容如下:(1)针对模糊时间序列模型处理数据时的不确定性和缺乏客观性,提出基于密度峰值算法的模糊时间序列模型,并将其应用于股票价格的预测中。(2)由于人工鱼群算法能够获取全局最优解,克服K均值算法对初始化聚类中心敏感且易陷入局部极值的问题,提出基于核函数人工鱼群的K均值算法,将其应用于股票市场的聚类分析中。(3)针对股票数据高维性和复杂性的特点,提出基于因子分析法和OPTICS-Plus算法的股票分类模型,实现对股票市场的分类。该算法有效地消除数据的冗余性,提高聚类的性能和收敛速度。(4)鉴于上市公司财务数据的高维性和冗余性等特点,提出基于Lasso方法和Logistic回归的上市公司财务预警模型,判断上市公司财务状况是否发生危机,达到预警的效果。本文应用数据挖掘中的分类和聚类算法预测我国股票市场的价格,对股票市场分类,以及预警上市公司经营状况。仿真实验结果表明,本文提出的方法能够较为有效地分析和预测我国股市,帮助投资者合理地做出决策。
[Abstract]:As an important part of the securities industry, the stock market has attracted the attention of investors. It is of great practical and theoretical value to seek effective stock analysis methods to reduce the risk of investors. Because the stock market is influenced by many factors, such as the intrinsic value of the stock, market factors, political factors and macroeconomic operating conditions, there are no definite rules among the various factors, and the stock market produces a large amount of data every day. All of these bring some difficulties to stock market research. Data mining combines database, statistics and artificial intelligence. It can extract hidden valuable information from a large number of raw data. The characteristics of stock market determine the application of data mining to the analysis and prediction of stock market. Through collecting and arranging stock index and financial data of listed company, using the classification and clustering method in data mining, this paper puts forward a series of improved algorithms for different problems. The main contents of the analysis and prediction of Chinese stock market are as follows: (1) aiming at the uncertainty and lack of objectivity in processing data in fuzzy time series model, a fuzzy time series model based on density peak algorithm is proposed. The artificial fish swarm algorithm can obtain the global optimal solution, and overcome the problem that K-means algorithm is sensitive to initialization clustering center and easily fall into local extremum. This paper presents a K-means algorithm based on kernel function artificial fish swarm, and applies it to the clustering analysis of stock market. Aiming at the characteristics of high dimension and complexity of stock data, a stock classification model based on factor analysis and OPTICS-Plus algorithm is proposed. The algorithm effectively eliminates the redundancy of data, improves the performance and convergence speed of clustering. (4) in view of the characteristics of high dimension and redundancy of financial data of listed companies, this algorithm can effectively eliminate the redundancy of data, improve the performance of clustering and speed of convergence. This paper puts forward a financial early warning model of listed companies based on Lasso method and Logistic regression to judge whether the financial situation of listed companies is in crisis and achieve the effect of early warning. In this paper, the classification and clustering algorithm in data mining is used to predict the price of stock market in China. The simulation results show that the method proposed in this paper can effectively analyze and predict the stock market in China and help investors to make reasonable decisions.
【学位授予单位】：吉林财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;F832.51

【参考文献】