Trading strategy prediction model based on quadratic programming and XGBoost

Abstract

In this paper, the quadratic programming is established on the basis of mean-variance evaluation method, and the commission is taken into account to achieve the effect of large investment return and small risk. The error analysis between the predicted data and the actual data is made. Then, the obtained data are introduced into the quadratic programming model to constrain the constraint conditions, and the percentage of daily investment in gold and bitcoin in 5 years is solved, which is the trading strategy. XGBoost regression model and BP neural network model are used for prediction. The error analysis of the two strategies is carried out, and the XGBoost model with high prediction accuracy is selected. The prediction accuracy can prove that this strategy is the best strategy.