Font Size: a A A

The High Frequency Quantization Strategy Research Based On Intraday Momentum Effect

Posted on:2020-01-16Degree:MasterType:Thesis
Country:ChinaCandidate:Z Z LuFull Text:PDF
GTID:2370330590487955Subject:Finance
Abstract/Summary:PDF Full Text Request
Quantitative momentum strategy in this paper is based on momentum effect,it means the change of stock price will keep the original trend for a period of time.With the further study of momentum effect by scholars,we find that there is significant momentum effect not only in mature European and American markets,but also in many emerging markets in the empirical study.This paper mainly studies how to combine the momentum effect with machine learning algorithm to build a machine learning-based model to predict stock prices.Through this model,we can get the effective buy signals from the “HS 300” stock index,and according to the effective signal,trading in the top ten stocks of 300 component stocks in “HS 300” based on momentum effect,then we can get the excess returns that beyond the market.Combining traditional financial theory with machine learning enriches and deepens the existing quantitative investment strategy,and provides a good idea and reference for the introduction of other financial theories.This paper chooses data of “hs 300” stock index from January 2008 to December 2018 as data samples,and 75% of the data samples are used as training set,that is,the data from 2008 to the March of 2016 are used as model parameter training samples,and use the remaining 25% of the data samples as test sets,that is the data from the April of 2016 to the end of 2018 are used as test set data outside training set samples.The process of model building is mainly divided into five parts: data classification,missing data processing,model building,model parameter optimization and result analysis.The results show that there are significant differences in the cumulative return rates in test sets within the holding period(one day)under different effect periods.When the effect formation period is within a day,the strategy can not achieve positive returns;When the effect period is one week,the strategy can achieve a three years' cumulative yield of 218.52%;When the effect period is two weeks,the strategy can achieve a three years' cumulative yield of 187.32%;When the effect period is four weeks,the strategy can achieve a three years' cumulative yield of 163.07%.The conclusion of this paper enriches the domestic research field of market momentum effect,and has a profound reference significance for relevant market supervisors and investors.
Keywords/Search Tags:momentum effect, machine learning, quantitative investment
PDF Full Text Request
Related items