Fluctuation Trend Prediction and Investment Allocation Optimization of Risk Assets and Safe-haven Assets

Abstract

Safe-haven assets are a safe and effective value storage and risk hedging tool in the period of market turbulence, while risk assets show multiple characteristics such as the coexistence of high risk and high return, great variability, strong volatility and so on. Taking gold and bitcoin, two typical safe-haven and risk assets, as examples, this paper constructs the ARIMA-XGBoost joint prediction model and predicts the future fluctuation trend of gold and bitcoin; At the same time, the prediction results are used to optimize the parameter allocation of the mean variance model, and the effective frontier of the portfolio is calculated under different constraints. The results show that the RMSE of ARIMA-Xgboost model is 5.3 and 83.6 respectively, and the MAPE is 0.35% and 0.80% respectively; The efficient allocation frontier of the portfolio is its Pareto optimal solution, and when the allocation proportion of a single asset is limited, the overall yield of the portfolio is significantly reduced, but it is better than the result of equal weight allocation; ARIMA-Xgboost model has high prediction accuracy, good stability and strong self-learning and self-adaptive ability, which can provide a certain reference for investors or salespeople to make investment decisions.