Prediction Of CSI 300 Index Trend Based On Random Forest Optimization

With the emergence and development of artificial intelligence,the "AI revo-lution" has also entered the financial sector,which has entered the era of big data.On the basis of financial big data,this paper fully mines the financial data,Using machine learning or deep learning to build the connection between the past and historical financial data,and then makes corresponding results through the latest financial data.Through these results,investors can conduct effective analysis and research on the financial market and reduce their invest-ment risks.Quantitative investment refers to the use of big financial data to fully mine the financial data,build the connection between the past historical financial data,use various financial derivatives,adjust the investment strategy,reduce the investment risk of investors,and provide good returns.Predicting the stock market has been studied for decades,but given its inherent complexity,dynamics and confusion,it has proved to be a very difficult task.The number of variables and sources of information to consider is enormous,making the task of predicting future stock price movements very difficult.The possibility of such a feat has been debated in the scientific community for decades.This paper uses the trading data of CSI 300 index to mine and screen out the characteristics that influence the price trend of the index,according to the characteristics of financial data,in a Random Forest model,(Random Forest,RF)as the foundation,combined with Genetic Algorithm(Genetic Algorithm,GA),GA-B-RF model is set up,build up their inner link to predict,the CSI 300 index future daily movement type,and further,using the data of five minutes,five minutes for predicting the trend for the future.The application of machine learning model to quantitative investment in domestic stock market is studied empirically.Based on random forest model applied to the forecast problem of the stock index and the direction,to sum up the experiences of past research and selection of the common technical indicators,on the basis of using MDI,MDA for feature selection,in order to solve the influence of the substitution effect,joined SFI to optimize the feature selection,and the optimization model of the input feature combination.Since financial data are not distributed independently,a new classi-fier model is constructed based on random forest based on the idea of integrated learning algorithm,and the trend of stock rise and fall is studied and predicted.A ga-b-rf prediction model with strong generalization ability was established by op-timizing the model parameters with genetic algorithm.At the same time,for the random forest model optimized by grid search and the random forest model op-timized by genetic algorithm,and compared with the GA-B-RF model,and the following conclusions were obtained:Using single feature importance(SFI)to supplement the MDI and MDA,the importance of characteristic,filtered to fea-ture combination,build Technical indexes to predict the direction of the Shanghai and shenzhen 300 index fall combination,reduce the influence of the substitution effect,and avoid abandoned not important characteristics,improve the prediction performance of the model and simplified model complexity.The Random Forest optimized by genetic algorithm has better performance than grid search,which indicates that genetic algorithm is easier to obtain the optimal solution.The GA-B-RF model can solve the overfitting problem of machine learning in the financial field to a certain extent,and has a good prediction effect on the trend predic-tion of CSI 300 index.The prediction accuracy reaches 60.43%on daily data and 66.67%on 5-minute data,which is higher than the traditional random forest model.