The Expansion Of The Quantitative Multi-factor Model Based On Artificial Intelligence And Its Application In The Chinese Stock Market

Quantitative investment is mainly based on mathematical methods to complete the decision-making and implementation of investment,which is often accompanied by the participation of computer technology.The theory,strategy and practice of quantitative investment have developed for many years in the mature overseas investment market.With the rise of artificial intelligence technology in recent years,a variety of new technologies,new models and high-performance computers are more and more closely combined with quantitative investment,and the related concepts and research are gradually paid more and more attention.Quantitative investment has many characteristics: first,it makes investment decisions strictly according to mathematical results.Decision making in quantitative investment is based on the results of the model,not the feelings of investors.Quantitative investment overcomes these shortcomings well.Every investment decision and all the reasons for action are all based on rigorous mathematical calculation results.Secondly,quantitative investment can achieve efficient and rigorous processing and analysis of data.Facing the explosive growth of information in today's information age,traditional investment methods are difficult to achieve comprehensive,meticulous and logical analysis and processing.The model of quantitative investment can deal with the information of thousands of investment objects very quickly.The latest machine learning algorithm can clear the relationship among various complex information and build the evaluation system.It can be said that the quantitative investment based on information technology can see the information that is difficult to see in the traditional investment way.Finally,the quantitative investment can explore the value from the financial data.From the initial statistical arbitrage,the more statistical principle is to find the price difference between the stocks with the same nature.Then,based on the probability and historical data mining model,quantitative investment uses mathematical knowledge to establish an investment method completely different from the traditional qualitative investment method,without the use of stock based nature analysis.In this paper,the multi factor model of quantitative investment strategy is expanded with the aid of artificial intelligence technology.The main work is as follows:First of all,quantitative investment strategy plays a leading role in the field of investment.The core of quantitative investment strategy is to allocate the portfolio according to the relationship between factors,returns and risks.Therefore,the core toimprove the efficiency of asset allocation lies in the continuous improvement of the factor model.This thesis proposes that using Elman neural network to predict the factors in the multi factor model can achieve better results intersecting with the traditional quantitative multi factor model.In this paper,taking the data of eight quarters in 2017~2018 as samples,we first test the effect of the multi factor model,and the results show that the multi factor model in the China's securities market has a significant effect.At the same time,the comparative experiments show that the factors in the future are better than those in the current period to predict the future return of assets.The traditional linear regression model needs more data to predict the factors in the future,which must meet the conditions of stability or cointegration.The test in this thesis shows that the factor data can not meet the relevant requirements.Therefore,this thesis proposes to use the nonlinear Elman neural network to predict the future trend of the factors.The experimental results show that the factors predicted by Elman neural network can play a certain role,and if the prediction factors are combined with the original multi factor model,it can obtain a significant better effect than the original model.This method expands the way of asset allocation and improves the asset allocation Efficiency.Secondly,this thesis studies the application of macroeconomic factors in the multi factor stock selection model.In the traditional multi factor model,we must first rely on Fama Macbeth regression to test the effectiveness of the factors.When the factors can not complete the effectiveness test,we think that under the traditional framework,we can not establish the relationship between the factors and the stock return,and then the macroeconomic factors can not be applied to the multi Factor stock selection model.However,the relationship between macroeconomic factors and stock return is in line with economic research.Therefore,in order to better describe the relationship between the two,this thesis proposes to use the non-linear model based on neural network to study the relationship between macroeconomic factors and stock return when the linear model can not establish the relationship between the two.Experiments show that the relationship between macroeconomic factors and stock return has been added The neural network model of macroeconomic factors can better fit the stock return.Furthermore,this thesis builds a multi factor stock selection model based on the prediction results of the neural network model.The experiment shows that the performance of the portfolio selected by the neural network model is significantly better than that of the traditional method in terms of the returnperformance,and the model added with the macro-economic multi factor model is better In addition,it proves the influence of macroeconomic factors on stock return.Finally,smart beta investment strategy is a kind of investment strategy based on the combination of alpha strategy and beta strategy.Its core is to change the weight determination method of the index while tracking the index.The traditional index and beta strategy are based on the market value,trading volume and other weights of component stocks.Smart beta strategy is based on alpha factor.Firstly,according to the construction method of smart beta strategy,according to the company's quality factor and equal weight method,this thesis constructs the index tracked,and compares the performance of two weight determination methods under smart beta strategy.It can be seen from the comparison that different indexes are suitable for different weight determination methods.Therefore,this thesis proposes a method based on the classification function of artificial intelligence algorithm to predict the suitable weight of the index.Then,BP neural network and linear layer model are verified respectively for the prediction of index suitable weight method.The results show that artificial intelligence algorithm plays a certain role in determining the weight of sum when classifying different indexes.