Research On Automatic Investment Method Based On Genetic Algorithm

After more than 30 years of rapid development of China's financial market,the system is becoming more and more perfect,and the market transactions are becoming more and more active.With the increasing application of artificial intelligence technology in the financial market,how to make profits from the financial market through automatic investment method instead of human is attracting more and more attention.Based on the quantitative engine,the automatic investment method executes different investment strategies on the perceptive quantitative data.Automatic investment engine is a necessary tool for the research of automatic investment method.However,common quantitative investment engines are characterized by commercialization,non-open source,low degree of freedom and great limitations,which make it difficult to serve the automatic investment research for Chinese stock market.Aiming at the above problems,this paper firstly constructs a quantitative knowledge base with diverse contents,accurate and complete data,and provides data support for the research.Then,a new quantitative investment backtracking test engine was built based on China's stock market trading rules to provide an experimental platform for the research.Genetic algorithm is often used in the super-parameter optimization and multi-factor selection of strategy model,but few studies have directly applied genetic algorithm to the optimization of multi-factor strategy.With natural parallelism,genetic algorithm can obtain better results with a small amount of data and a small amount of calculation cost,which is very suitable for automatic investment strategy research under a small amount of short-term stock market data.This paper will study how to use genetic algorithm to optimize a multiple factor screening strategy,so as to achieve an end-to-end automatic investment framework.The work is mainly divided into the following several aspects: in terms of strategy initialization,common initialization method based on uniform distribution due to the strategy search space is large and sparse,will lead to initialization set will appear a large number of invalid strategy problems,unable to provide effective initial population genetic algorithm.This paper proposes a strategy initialization method based on random portfolio,which solves the problem of a large number of invalid strategies in the initialization set.In terms of strategy coding,this paper proposes a multi-factor masking coding method,which enables the genetic algorithm to automatically screen effective factors.In terms of strategy fitness evaluation,the quantitative investment engine is combined to carry out backtracking test on the strategy,and the information ratio of the strategy during the investment period is used as the fitness value of the strategy,which help the genetic algorithm find the optimization direction.In addition,this paper also carefully designed the process of strategy selection,strategy crossover,strategy variation and so on.The improved genetic algorithm is adopted to optimize the multi-factor screening strategy,which achieved 13.28% excess return during the out of sample investment period from June 2 to August 1,2020,exceeding the CSI300 index in the same period and surpassing other similar strategies.Therefore,the automatic investment method combined with genetic algorithm has certain utility.