Inverted Error Combination Optimization Algorithm And Its Application In Financial Index Prediction

At present,the stock market has become one of the main markets for Chinese people to invest and finance.It is very important to predict the future trend of the stock market,which is the key to obtain excess returns from investment and finance.The changes of stock prices in the Chinese stock market can be well reflected in the CSI 300 Index(namely Shanghai and Shenzhen 300 index)and CSI 500 Index(namely China Securities Small-cap 500 Index),in which the mainstream investment returns on the market and the whole development of small and medium-sized market value of the company can be respectively represented in the CSI 300 Index and CSI 500 Index,too.In this paper,a combinatorial optimization algorithm of inversion error is proposed,and a combinatorial model is established by weighting the inversion error method for regression prediction.Grid search algorithm and genetic algorithm optimization scheme are used to respectively optimizing parameters for the combination model of extreme gradient tree XGBoost model and support vector regression machine model(namely SVR model)based on the radial basis RBF kernel as well as the optimal parameters are found by comparing the experiment error.The revised forecast error is weighted by the inverted error method so that the combination model is presented in the paper.When combining multiple models,not only the problem of weight setting should be considered,but also the self-prediction ability of a single model should be considered,so that the combination model with the weight slightly biased to the strong prediction model should be selected.The daily opening price of CSI 300 Index and CSI 500 Index can be predicted by the combination model as well as its value of practical application is showed in the experimental results.