Research On Alpha Arbitrage Of Support Vector Machine Timing

With the development of the economy and the accumulation of national wealth,the problem of how to make the assets appreciation attracts more and more people's attention.On the one hand,due to the fluctuation of domestic and foreign economy and the imperfection of financial market rules,the volatility of financial market in China has occurred sometime.Therefore,Alpha strategy is favored by investors in recent years as an investment way which can avoid systemic risk.On the other hand,there are obvious shortcomings in the traditional Alpha strategy,long-term bilateral risk exposure,and once the financial market volatility,cointegration relationship between stock index futures and spot CSI 300 index suddenly weakened.Then the traditional bilateral Alpha Strategy not only can not hedge the system risk,but will increase the degree of bilateral losses.Taking into account of the facts that the market conditions known in the case of Alpha arbitrage strategy is often a better performance,this paper purposes on how to predict market trends to reduce the holdings of bilateral positions,narrow the exposure and expand the profitability,and make up for the traditional Alpha strategy's flaws.The first is predicting the market trend.This paper chooses the opening and closing price,the highest price,the lowest price,the turnover,the volume,the fluctuation rate and the amplitude of the Shanghai and Shenzhen 300 index from January 1,2006 to December 31,2015,and uses the support in the machine learning Vector machine technology to forecast the trend of the next five trading days of the CSI 300 index.And the cross validation,genetic algorithm and particle swarm algorithm are used to improve the parameter estimation problem.Based on the optimization of the parameters,the performance of the four kernel functions,such as the linear kernel function,the polynomial kernel function,the radial basis function and the Sigmoid kernel function,are compared based on the mean square error and the square correlation coefficient index.Cross validation and genetic algorithm Optimization of the parameters are suitable for choosing the radial basis of the kernel function to predict the market trend,particle swarm optimization algorithm is better to select the linear kernel function to predict the market trend.The second is studying the method of construction of Alpha arbitrage combination.This paper chooses data of 129 stocks in Shenzhen stock market from January 1,2006 to December 31,2010.The data includes the closing price,the total market value,net profit,operating income,volume,turnover rate.Then the six indicators are converted into yield,price,Market sales rate,volatility,turnover,liquidity,and using the panel binary selection model to look for the factors that has obvious significance for stocks beyond the market index.The study finds that volatility,turnover and price-earnings ratio play a significant role.Therefore,scoring the 129 stock according to these the factors and choosing the top 30 stocks to combilate alpha arbitrage.The strategy is that if the stock will move up,the alpha arbitrage only holds a long stock portfolio;if it falls,it also holds a long stock portfolio and a short stock index futures.The last third is the empirical study of arbitrage effect.This article uses data from January 1,2011 to June 31,2014 and July 1,2014 to December 31,2015(the former fluctuates little and the latter fluctuates violently).This paper compare the arbitrage effect of the method and the traditional method from multiple perspectives.The results show that the alpha arbitrage with predicting market trends is superior to the traditional method,the market strength index performs well when the market volatility is not violent,and vice versa support vector machine effect is good.In addition,this paper only considers only one kind of kernel function methods,if you can consider the integrated learning method to summerize the results of different kernel,you may achieve better arbitrage effect.