Research On Intraday Momentum Strategy Based On Machine Learning

With the rapid development of technology,high-frequency quantification has be-come a hotspot in the research of global financial markets,and the intraday momentum effect has become the focus of the majority of scholars.Therefore,this article mainly uses advanced algorithms and theories to conduct an in-depth study of the intraday momentum effect in the Chinese market,that is,the momentum that forms within the day and ends within the day.First of all,this article conducts a traditional intraday momentum effect test on stock indexes and stock index futures,which uses the jump information at the opening and the first half-hour yield information to predict the rise and fall of the last half hour.Secondly,since the intraday momentum effect is mainly caused by changes in trading volume and investment sentiment,this article combines the intraday momentum effect with changes in investment sentiment on different trading days.The experimental results show that investors are in an excited investment state on Monday.Investment sentiment was sluggish on Friday,and the intraday momentum effect was significant on Monday and not significant on Friday.Then,based on the reason that interest rate fluctuations will cause changes in in-vestor sentiment and trading volume,this paper combines interest rate fluctuations with intraday momentum effects to further study the intraday momentum effects of stock index futures under interest rate fluctuations.The experimental results show that when the fluctuation of the inter-bank buy-out repo rate exceeds the range of 10%of its 5-day average online and offline,investor sentiment is significantly higher,the trading volume of stock index futures increases rapidly,and the intraday momentum effect also increases accordingly.Finally,the intraday momentum effect of stock index futures is not obvious,because of the special trading mechanism of A shares,the price trend of stock index futures shows a non-linear and non-stationary characteristic.Therefore,this paper uses three machine learning models(SVM,Naive Bayes,XGBoost)to study the intraday momen-tum effect of the Shanghai and Shenzhen 300 stock index futures.Among the three models,the empirical results of the intraday momentum strategy under the XGboost model are the most significant.The accuracy of the strategy is 55.8%,the annualized return rate is 18.2%,the Sharpe ratio is 2.45,and the maximum retracement is 7.62%.The intraday momentum effect under the XGboost model not only can achieve excel-lent benefits when considering handling fees and slippage,but more importantly has real application value.