Portfolio Risk Investment Strategy Model based on ARIMA and LSTM

Abstract

Market trading is often fraught with uncertainty. We build a model to predict the future market movements of gold and bitcoin prices based on historical data. We also design a reasonable trading strategy that achieves the expected returns. First, we build a market forecasting model based on ARIMA and LSTM neural network to obtain a set of price change series that can reflect both the future trend and observe the recent increase to support the design of a market trading strategy. Next, we designed a portfolio investment model that combines return and risk. Combining the commission at the time of trading with the return on investment avoids the impact of considering commissions at the time of decision making. Each held product is decided separately in the investment decision, and the weighting of each investment is adjusted by planning the weighting of each investment with the expected risk. The acceptance of different risks, the R-value of ARIMA prediction results, and the stage evaluation of LSTM are integrated to adjust the weighting of risks, which finally gives three different risky portfolios of low, medium, and high.