Improvement And Application Research Of AdaBoost Algorithm In Quantitative Investment

As a new investing method,quantitative investment has achieved ever-increasing market size by its consistent performance.Machine learning is a subfield of artificial intelligence concerned with techniques that analysis financial unstructured information effectively by its strong data handling capability and self-learning ability.As one of classical machine learning algorithms,AdaBoost algorithm has become focus and got much in-depth theoretical research and widerranging application.However,AdaBoost algorithm's research on quantitative investment is seldom involved.This paper applies AdaBoost algorithm to quantitative trading and improves algorithm according to financial market information.Firstly,the background of machine learning applications in quantitative investment filed and its basic theoretical methods are introduced.Then Random Forest algorithm's feature selection principle and algorithm design are introduced for the purpose of selecting optimal data set to algorithm model.Furthermore,in terms of enhancing applicability,Ada Boost algorithm's improvements includs initial sample in exponential decay weights,measuring market fluctuations by Hurst exponent and adjusting the weight of weak classifiers based on soft margins.Lastly,a set of AdaBoost-EHS based quantitative trading strategy is established,which indicates the improved algorithm has obvious ensemble effectiveness.And the trading strategy has passed commodity futures' universality test with strong profitability.The research on Ada Boost algorithm's improvement can broaden trading strategy design in futures markets,which also provide certain reference significance for machine learning methods' application in quantitative investment.Whereas,more work is needed to explore in parameter optimization,margin distribution theory,strategy design and so on.