基于双T-SV模型下支持向量机回归的量化策略研究

发布时间：2018-04-18 16:12

  本文选题：量化投资 + 支持向量机　； 参考：《西南交通大学》2017年硕士论文

【摘要】：量化投资是一种的更理想的投资方式,它以数据为基础、以模型为核心、以程序化交易为手段,且具有总体收益稳定、持仓时间较短、交易标的较多等特点。同一般的传统投资方法相比较,量化投资具有如下的优点。首先,量化投资依赖于更加客观的投资逻辑。量化投资从决策产生及策略执行都是通过计算机程序来实现,这样可以消除人为情绪所造成的不良后果。另外,量化投资的指令更加精确、交易更加快速。量化投资在利用计算机收集、处理历史数据的特性下,能够更加高效、更加全面地分析数据,不错过每一个可能盈利的机会。在中国,量化投资还处于发展起步阶段,整个金融投资领域,量化投资所占的比重甚至不到十分之一,说明量化投资在国内的股票市场仍有非常广阔的发展前景。随着中国金融市场的不断完善和发展,不断地推进金融改革和金融创新,量化投资在中国金融市场这片乐土上必将茁壮成长。整个投资过程都有量化投资技术的身影,其中包括量化选股、风险控制、算法交易等,本文针对股票收益率和波动的预测建立量化选股策略。首先,基于两个随机扰动都服从尖峰厚尾的T分布的SV模型的基础上,构建了双T-SV模型。基于传统的先验分布假设,推导了双T-SV模型的MCMC估计过程,并将其用于HS300指数的实证分析。通过与传统的SV模型簇的DIC准则对比,证实了双T-SV模型能够更加准确刻画我国金融收益率波动时变性、聚集性。针对股票的收益率,本文通过对六大类因子的选取,主成分分析预处理,利用前6个主成分来作为输入,以每只股票的后五个交易日的累积收益作为输出,建立支持向量机回归模型,基于技术面、成长面等因子的支持向量机回归模型成功预测股票的五日的累积收益率。结合支持向量机预测与双T-SV模型,构建了量化投资策略。该投资策略以双T-SV模型预测的收益率波动σ与支持向量机的预测收益率r相结合来进行选股,在HS300股票市场中选出r-λ*σ0(其中λ为风险厌恶因子,0≤λ≤1)的股票加入备选股池,并且在备选股池中选出r-λ*σ值最大的50只股票加入买入集,每五日调仓。根据量化策略在HS300股票市场上进行回测的结果,其年化收益率可超过33%,累计收益达到117.8%,并且夏普比率达到0.83。在裸多的情况下,其资金收益率远远超过沪深300指数,说明了策略的优越性。
[Abstract]:Quantitative investment is a more ideal way of investment. It is based on data, takes model as the core, takes programmed transaction as the means, and has the characteristics of stable overall income, short position time and many trading targets.Compared with conventional investment methods, quantitative investment has the following advantages.First, quantitative investment depends on more objective investment logic.Quantitative investment is realized by computer program from decision generation and strategy execution, which can eliminate the adverse consequences caused by artificial emotion.In addition, quantitative investment orders are more accurate, trading faster.Under the characteristic of using computer to collect and process historical data, quantitative investment can analyze the data more efficiently and comprehensively, and not miss every possible profit opportunity.In China, quantitative investment is still in the initial stage of development. The proportion of quantitative investment in the whole field of financial investment is less than 1/10, which indicates that quantitative investment still has a very broad development prospect in the domestic stock market.With the continuous improvement and development of China's financial market, financial reform and financial innovation are constantly promoted, and quantitative investment is bound to thrive in this happy land of China's financial market.There are quantitative investment techniques in the whole investment process, including quantitative stock selection, risk control, algorithm trading, etc. In this paper, quantitative stock selection strategy is established for stock return and volatility prediction.Firstly, based on the SV model of T distribution with two random disturbances, a double T-SV model is constructed.Based on the traditional prior distribution hypothesis, the MCMC estimation process of the double T-SV model is derived and applied to the empirical analysis of the HS300 index.By comparing with the DIC criterion of the traditional SV model cluster, it is proved that the double T-SV model can more accurately describe the volatility and aggregation of the financial yield in China.According to the return rate of stock, this paper uses the first six principal components as input and the cumulative income of the last five trading days of each stock as the output through the selection of six categories of factors and the preprocessing of principal component analysis.The support vector machine regression model is established, and the support vector machine regression model based on technical and growth factors is used to successfully predict the cumulative return rate of stocks for five days.Combining support vector machine prediction with double T-SV model, a quantitative investment strategy is constructed.The investment strategy combines the return fluctuation 蟽 predicted by double T-SV model with the predicted return rate r of support vector machine to select the stocks with r- 位 * 蟽 0 (where 位 is risk aversion factor 0 鈮，

本文编号：1769099