The Analysis Of Chinese Stock Price Influence Factors Based On Lasso

Since the reform and opening up, China’s economy has achieved rapid development, China’s gross domestic into production （GDP） increaseing from18,774.00 （million） in 1990 to 634,043.40 （million） in 2014. At the same time, China’s securities market has got more comprehensive growth and improvement of socio-economic development to make a contribution can not be ignored. Although the establishment of the stock has been close to four hundred years of history, the Chinese stock market started relatively late and even the world, but the growth rate and scale should not be underestimated. First, from the deal, according to China Securities Regulatory Commission statistics, in 2013 China’s stock turnover accumulated 468,728.60 （billion） in 2014 grew 58.71 percent over the previous year, while in 2015 at the end of 12 totaled 2,550,538.30 （million） compared with 2014 growth of more than 200%. Secondly, trading activity, the Shanghai Stock Exchange alone, the total turnover in 2015 amounted to 1,024,856,267.14 （million shares）, and the highest daily turnover in 2015 reached 8,607,174.87 （million shares）. So it can be said that the securities in the stock market can be used to reflect China’s macro economy is running a "barometer", it has to be reckoned with in terms of the role of the modern market economy.For China, the stock market launched a late, only 20 years of history of the national stock, China’s stock market, there are still many variables. December 2014 and January 2015, the market is almost spent in severe shock in 2015, China’s stock market is doomed to crash from mad cow memorable year. From the start to the formation of mad cow bull, to the outbreak of the stock market crash of liquidity completely lost, there appeared six days fell 1,000 points and 3000 down in 2 months, all completed in just six months time, and as a dream magic.Securities in the stock market reflects China’s macro economy is running a "barometer", so the market price of the stock research has great significance.This paper aims to study the Shanghai Composite risk factors. In particular, for the national economy and the stock market situation, initially singled out 28 candidate factors, these factors reflect the cover price index, the economic sentiment index, monetary policy, interest rate policy, fiscal policy, foreign trade and foreign exchange trading activities reflect the stock market itself trading volume and impact of the situation in China and the world average index of the Dow Jones industrial.Multiple linear regression model to study financial securities stock market have a good practical value, can be used to identify and explain the impact of multiple risk factors on outcome variables. Before creating a multiple linear regression model, for a detailed interpretation of the dependent variable, we tend to collect in advance as much as possible the impact factor, whether transverse or longitudinal impact influence, whether directly or indirectly affected by the impact, we want to add to the model to better explain how the dependent variables that influence these factors. When too many factors, on the one hand, there may be multiple linear relationship between these factors, the conventional least squares regression method is not applicable; on the other hand, too many factors not conducive to our interpretation of the model. At this time, the ridge regression （ridge regression） by using L₂ punishment of regression coefficient model is compressed in order to solve the problem of multicollinearity, resulting model has good stability. But the final ridge regression regression coefficients can not be compressed to 0, variable selection can not be performed, resulting in the model is difficult to explain. But we want to pick out has a significant influence on the dependent variable factor from a small factor in the alternative, at this time, we usually use the variable selection method. The traditional variable selection methods include forward stepwise regression （forward regression）, backward stepwise regression （backward regression） and the combination of the two regression （stepwise regression）, these methods are mostly based on some information criteria, including AIC and BIC. The information criterion is a combinatorial optimization problem, when a high number of dimensions, there will be NP-hard problem that calculation time will increase with the number of dimensions showing exponential growth, and the statistical properties of these methods is not clear. Recently, Tibshirani Lasso method proposed by the coefficients L₁ penalty, some very small coefficients can be extracted directly from 0 to automatically achieve the purpose of variable selection, and at the same time the estimated regression coefficients of significant variables, but the completion of the selection and We estimate, so the calculation is extremely simple. Many recent studies （Fan & Li2001 and Fan & Peng2004） show Lasso method has consistency and asymptotic normality.Combining multiple regression models and methods to identify and estimate the Shanghai Composite Lasso risk factors. In particular, this paper selects the monthly data from the Shanghai Composite Index and 28 risk factors 2010.01 months to 2015.12 of the month. First, we log-transformed data, the Shanghai Composite Index, to reduce its heteroscedasticity, in addition to the number of data converted by modeling a better explanation. Secondly, we have the data unit root test --ADF test to examine the stability variables. Next, in order to explore the long-term equilibrium relationship between the dependent variable and the explanatory variables, the paper dependent and independent variables were EG cointegration test found that both co-integration relationship. Then we continue to examine the Granger causality is interpreted and explanatory variables between the final results show that the explanatory variables are stationary after the Granger cause of the Shanghai Composite Index monthly returns. Finally, we use standardized data Lasso model.For comparison, the fourth part of this article also gives three criteria based AIC stepwise regression method estimates the model results obtained under each method and compared. Both methods show than elected Shanghai Composite Index monthly volume growth, consumer satisfaction index an incremental revenue growth as well as the Dow Jones index monthly returns, and on the Shanghai Composite Index monthly returns significantly affected the degree of the top two are the Shanghai Composite index and the Dow Jones Industrial average monthly volume growth in monthly returns, which is the volume and rate of return is the fact that the stock market is the most important of the two indicators is understandable.The results also show Lasso than obtained a better model, in particular in:.1 Lasso to have greater explanatory, Lasso is selected five variables, stepwise regression elected eight variables, and exchange rate, stepwise regression is selected aspects of the Hong Kong dollar exchange rate, and Lasso selected is the most representative aspects of the dollar exchange rate; 2 with respect to the stepwise regression analysis, the results Lasso has a smaller prediction error, fitting effect better.In this paper, the Shanghai Composite Index display monthly volume growth, consumer satisfaction index an incremental, incremental dollar closing rate, revenue growth as well as the Dow Jones index monthly returns five variables have a significant impact on the Shanghai Composite Index monthly returns.