传递函数模型在股市分析中的应用

发布时间：2018-08-27 08:59

【摘要】：随着中国股票市场不断发展和完善，时间序列分析用于股市的研究越来越多。传递函数模型在经济、工业以及工程领域方面应用较广泛，在股市应用仍然少见。本文将传递函数模型应用于股市分析，主要利用价量的互相关函数检验，分析在牛市、熊市和平衡市下，上证综合指数价量关系。并与恒生指数、道·琼斯工业股价平均指数和上海个股中国石化进行比较。得出了上证综合指数在不同趋势下结果存在共性和差异，国内外指数、指数和个股在同一个趋势下结果存在共性和差异，，有价量存在同期关系，有也有价量互相反馈关系，还有价导致量或量导致价。再对价量建立传递函数模型，其优点是模型的传递函数部分表达出价量的关系，噪声部分反映了输出序列自身的时序关系。对于输出时间序列，若与输入序列的影响越显著、影响总期数越多时，建立的传递函数模型会优于ARIMA模型。反之，影响不显著，或者影响总期数较少，建立ARIMA模型反而更优。对多变量传递函数模型的实证研究，本文先利用多元统计学中的逐步回归法，以及输出序列与输入序列的互相关函数，先剔除了不满足条件的输入序列，再建立模型。当多个输入序列之间存在高相关性的情况，本文尝试两种加权方法，构造新输入序列，其中根据输出序列和输入序列的互相关函数的比值得出权重而构造的新输入序列，与输出序列建立的模型最优。针对干预事件的出现，本文在第三章理论基础上提出了三种修改的传递函数模型，对其中输入变量带干预的传递函数模型进行了实证分析，以股指期货上市对股市的干预为例，实证结果表明在输入序列受到干预时，采用提出的修改模型优于原始的传递函数模型。
[Abstract]:With the development and perfection of Chinese stock market, time series analysis is more and more widely used in stock market research. Transfer function model is widely used in economy, industry and engineering, but it is still rare in stock market. In this paper, the transfer function model is applied to the stock market analysis, which mainly uses the cross-correlation function test of the price quantity to analyze the relationship between the price and quantity of the composite index in the bull market, the bear market and the equilibrium market. And compared with Hang Seng Index, Dow Jones Industrial average Index and Shanghai stock market Sinopec. It is concluded that there are commonness and differences in the results of the Shanghai Composite Index under different trends, and there are commonness and differences in the results of the domestic and foreign indexes, indices and individual stocks in the same trend, and there is a contemporaneous relationship between the price quantity and the mutual feedback relationship between the price and the quantity. And price leads to quantity or quantity to price. The advantage of the transfer function model is that the transfer function part of the model expresses the relation of the price and the noise part reflects the sequential relation of the output sequence itself. For the output time series, if the influence of the input sequence and the input sequence is more significant, the established transfer function model will be better than the ARIMA model. On the other hand, if the influence is not significant, or the total number of periods is less, it is better to establish ARIMA model. Based on the empirical study of multivariate transfer function model, this paper firstly uses the stepwise regression method in multivariate statistics and the cross-correlation function between output sequence and input sequence to eliminate the input sequence which does not satisfy the conditions, and then establishes the model. When there is a high correlation between multiple input sequences, this paper attempts two weighted methods to construct new input sequences, in which a new input sequence is constructed based on the ratio of the cross-correlation function of the output sequence and the input sequence. The model established with the output sequence is optimal. According to the emergence of intervention events, this paper puts forward three modified transfer function models on the basis of the third chapter theory, and makes an empirical analysis of the transfer function model with the intervention of input variables, taking the intervention of stock index futures listing on the stock market as an example. The empirical results show that the proposed modified model is superior to the original transfer function model when the input sequence is interfered.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F832.51

【参考文献】