Study On Portfolio About Dual Factors

Decision Tree Algorithm,Random Forest and Support Vector Machine model are the sub-branch areas of machine learning.They have been favored by various scholars and brokerage professional teams for many years,and they are applied in the field of quantitative investment.In this context,the construction of factors and the prediction of stock trends are topics that experts and scholars in this field are constantly pursuing.This paper hopes to innovate to make a break through on the base of the index system of existed articles,combined with relevant machine learning algorithms.This paper divides the stock selection process into two stages according to the different natures of the fundamental and technical indicators.During those stages,the fundamental indicators are used to judge the long-term status of the companies and the technical indicators to judge the choice of trading.Firstly,the first chapter describes the background and significance of the study,relevant literature review,research objects and research tools.The second chapter,the Two-Factor Model Built base on the portfolio construction of the two-factor model,focuses on the idea of portfolio strategy construction of the two-factor model,and introduces the two-factor model in detail.The third chapter conducts empirical research on the first factor(stage)of the two-factor model strategy proposed in this paper,namely the fundamental factor.The fourth chapter is the optimization and improvement of the previous section.On the basis of selecting target listed companies in the previous section,technical factors(stages)are added to improve the effect of stock selection strategy by using Machine Learning method.The fifth chapter summarizes the research content of this paper combined with the specific empirical situation.Secondly,this paper applies the Factor Analysis method and Decision Tree,Random Forest and Support Vector Machine methods to the two aspects of fundamental stock selection and technical stock selection(time selection)respectively.The results show that the trading strategy based on the dynamic Support Vector Machine algorithm can obtain the cumulative income of 3.29 times compared with the CSI 500 index and three times to All-Finger Consumption Index during the period and the cumulative income over the Full Index Consumption Index,and the corresponding cumulative return rate is 106 %.Compared with the Stock selected by Finance factors only,which's cumulative rate of yield is 23.5% during the first half-year period,the result has been greatly improved.The former's value is equivalent to 4.51 times of the latter.In addition,portfolios structured by technical indicators has better performance in terms of capital risk than unoptimized model.Lastly,based on the 1H stock market data during 2018,we will select several stocks in the Major Consumer Industry from the fundamental perspective and apply the technical indicators to forecast the Ups and Downs of stocks selected from the first phase,then the corresponding buy and sell operations will be conducted to build a portfolio during the period.It has vital significance for improving the diversity of stock selection investment strategies.