Research On High Frequency Data Trend Prediction Model Of Commodity Futures Based On LSTM

Quantitative investment is a new mode of active investment management.It has a history of more than fifty years in western countries,and only ten more years in China.In recent years,with the blowout development of machine learning and deep learning,more and more quantitative models based on machine learning and deep learning have been applied in the quantitative investment market.This paper analyses the researches and applications of machine learning and deep learning models in quantitative investment markets at home and abroad,and finds that they are mainly concentrated in the stock market and rarely in the futures market;Therefore,on the basis of introducing the related technology of time series prediction and analyzing the microstructure of the futures market,we select the high-frequency data of main contracts of rebar futures in commodity futures market from April to July 2017,and select its relevant features as input,establish a trend prediction model based on LSTM algorithm.The experimental results and back-test results show that the model can accurately predict the short-term fluctuation trend of rebar futures,which proves the applicability of LSTM algorithm in high frequency futures data.