Prediction Of Financial Product CSI 300 Based On SSA-LSTM LightGBM Model

In order to promote the healthy development of the modern financial product,ensure the sustainable and stable improvement of the national economy,avoid financial turbulence and avoid financial crisis,it is necessary to analyze and predict trend of financial products.Stock and fund have become important components of modern finance.Usually,the stock market reflects the economic and social conditions to some extent,and is known as the barometer of the economy.By analyzing and researching the stock market,the clue and information can be provided for financial products,providing scientific references for decision-making and consultation.Thereby better serving will provided for finance,economy and society.In this thesis Shanghai-Shenzhen 300 index(CSI 300),which is closely related to finance was selected for this research.The research object is to predict the CSI 300 Index.A SSA-LSTM-Light GBM prediction model based on the combination of least squares weight(OLS)was proposed and the entropy weight method(EW)is proposed too.Firstly,the four popular machine learning prediction models,namely BP model,Cat Boost model,LSTM model,and Light GBM model were established to predict the Shanghai and Shenzhen 300 index.The prediction data were compared and analyzed with the actual data.By comparison,the LSTM model and Light GBM model were selected as the excellent dominant models.Then,combining the characteristics of the respective models,we use Differential Evolution(DE),Firefly algorithm(FA),Particle Swarm Optimization(PSO)and Salp Swarm Algorithm(SSA)to optimize the LSTM and Light GBM model with hyperparameter and select the optimal model parameters.Finally,the OLS weighting and entropy weighting methods are used to fuse the prediction result of these two advantageous models and the final predicted data was obtained.The research results indicate that the combination prediction model can fully combine the advantages of different prediction models,achieving a high level of prediction accuracy and reliability.In the prediction of CSI 300 Index,the evaluation indicator RMSE in the optimal model is55.61,MSE is 3092.76,MAE is 42.19,and Goodness of fit is 0.98.When the tolerant error value is 5%,the prediction accuracy of the optimal model can reach 99%,which is significantly better than the single prediction model.The conclusion is that the optimized combination prediction model through parameter optimization has strong practicality and prediction accuracy,and can accurately predict the trend and change of the CSI 300 Index.This study can better provide decision-making basis for the financial products.