Research And Application Of Ouantitative Investment Decision Based On Machine Learning

Quantitative investment analyzes a large amount of financial data by computer,and makes trade decisions based on investment decision model.The model can update real-time data quickly,make rational decision,control gains and risk in investment,and grasp the market timing.With the rapid development of the Internet and big data technology,in the face of complex and volatile financial market and a variety of financial products and their derivatives,Investment through traditional quantitative investment strategy can no longer satisfied the demand for the huge number of market demand.In order to solve the problems,Some researchers and practitioners completing the transformation from traditional investment to quantitative investment,They're starting to make the transition to artificial intelligence quantitative investment.Machine learning,Deep learning and other algorithms are applied in the field of quantitative investment.The performance of the new quantitative investment method is more stable,and the market scale and market share are increasing.The new method is gaining the trust of more and more investors.In response to complex financial market signals,the signal decomposition technology are applied to the financial time series data.Separating effective financial signals from complex market noise,then mining valuable information from signal combinations of different frequencies.At the same time,in order to obtain trend information,neural network algorithm is used to predict the trend of time series.In order to build an effective quantitative investment decision model,select appropriate strategies for specific financial products and combine them with machine learning models.The CEEMDAN-LSTM quantitative investment decision model is established for suitable for stock index futures market with high return and low risk.The model combines signal decomposition technology,machine learning technology and CTA quantitative investment strategy.The model was established with the 1-minute market data of CSI 300 Stock Index Futures as the test data.In this model,CEEMDAN as a sequence decomposition module can decompose the fluctuations or trends of different scales in time sequence,producing high frequency noise sequence and low frequency trend sequence.Then,Predicting CSI 300 Stock Index Futures 'trends by LSTM neural network.The model was established by combining the trading rules of stock index futures market and quantitative investment strategy.The model with a higher Sharpe ratio and a smaller maximum retracement.In order to get a higher performance algorithm model,the advantages of different algorithms can be used to avoid the shortcomings of the model.