Optimal Solution for the Gold Bitcoin Portfolio Investment Model

Abstract

This topic is a portfolio investment problem with quantitative trading as the background. In order to solve this problem, three types of mathematical models are used in this paper, namely the prediction model, decision model, and risk assessment model. The first is the forecasting model. The paper applies three forecasting models: the grey system Grach (1, 1) forecasting model, the quadratic exponential smoothing forecasting model, and the time series BP-neural network forecasting model. The second is the decision-making model. The decision-making model in the paper is a constrained linear programming model. The objective function is to maximize the total revenue of the day. Finally, there is the risk assessment model. The quantitative investment and multi-factor models are used in the paper to calculate the standard deviation of the rate of return (conventional risk) of gold and Bitcoin respectively in a 30-day cycle, so as to achieve the purpose of quantifying risk, thus reflecting the relationship between gold and Bitcoin. The investment risk index of the two futures products of the currency is provided as a reference for investors. This paper also adjusts the parameters of the prediction model, such as adjusting the value of the number of neurons in the hidden layer of the BP-neural network, to compare the fitting effects corresponding to different parameters, to prove that the prediction model is an optimal solution; Give the decision-making model a certain disturbance, such as changing the definition of the objective function for the total return of the day, to reflect the performance of the forecasting model in dealing with the disturbance of special factors. After that, this paper also conducts a sensitivity analysis of the decision-making model. The specific method is to give a small disturbance to the decision-making model, such as changing its transaction cost, that is, the value of the commission rate a%, recording the final benefit of the decision-making model, and generating a chart to reflect the model. smoothness and sensitivity. Finally, this paper optimizes some models, such as optimizing the BP-neural network model by adaptively adjusting the learning rate and optimizing the linear programming decision model by adding the MACD information factor.