Empirical Study On Quantitative Investment Strategy Based On Bootstrap Method And Hidden Markov Model And Random Forest Model

In recent years,quantitative investment has drawn increasing investors' attention.Especially in China,the stock market and the futures and the options market have a very large number of underlying assets for trading and investment.Because machine learning itself matches the characteristics of quantitative investment,machine learning has gradually become the main research direction of quantitative investment.How to combine the advantages of machine learning with quantitative investment,and how to avoid the shortcomings of machine learning's over-fitting has become a major problem to be solved by many practitioners.This article mainly focuses on the domestic active commodity futures--iron ore futures,and designs a model with its continuous contract daily data.We extract features from price and volume data,and further test the effectiveness of excess returns and screen features which can describe the market status well.We then construct a ramdom forest model to predict the market ups and downs and finally form a quantitative trading strategy for iron ore futures.This paper first introduces the development of quantitative investment,machine learning model and other fields,and then briefly describes the theoretical basis of knowledge related to the empirical part.We then conducted an empirical study using the daily data of the main continuous iron ore contracts.The first step in the empirical study is to construct features that are sourced from the price and volume information of the iron ore and then test the validity of excess returns for each feature using a non-parametric bootstrap approach.Then,using the results of the previous screening,we use the unsupervised Hidden Markov Model to test which features can describe the status of market well.Subsequently,we constructed a model using indicators screened from the above steps to predict out-of-sample data.Finally,based on the forecast result,an iron ore quantitative investment strategy is formed.The empirical results of this paper shows that,the random forest model constructed in this paper can achieve a certain degree of accuracy in a short time scale,and can obtain higher returns and lower maximum drawdown from the perspective of quantitative strategy performance.The combination of machine learning and quantitative strategiescan yield steady excess returns.The research results of this paper have a certain application value,and follow-up research can also be based on this with further exploration.