Research on Quantitative Trading Decision Based on BP Neural Network and Circular Decision Model

Abstract

Market traders usually buy risky assets in order to achieve rapid asset appreciation. In this paper, we build a BP Neural Network Model (BPNNM) and Ci. color Decision Model (CDM) provides investors with optimal strategies by analyzing historical prices, risks, and other factors of gold and bitcoin. Firstly, we build a BP neural network to predict the closing price of gold and bitcoin. We make full use of all available data up to that day. Eight consecutive days of closing price are selected as training data (the closing prices of the first to seventh days are used as input data and the eighth day as output data). Accordingly, we utilize the BP neural networks to predict the next day's closing prices of gold and bitcoin, respectively, passing the reliability and validity tests (R2 > 0.99). Multi-objective nonlinear programming was established with the data predicted by the BP neural network. The objective is to maximize the next day's return and minimize the risk. Subsequently, we looped the multi-objective nonlinear programming daily to build a CDM which is solved to obtain the optimal strategy. The results show that we have an average annualized return of 172.73%.