Trading Method Based on ARIMA Model and Greedy Algorithm

To predict the price of a trading day, we construct ARIMA Model and Random Forest Model. To get the best ARIMA Model, we continuously change the parameter of the model. At the same time, we used the CART decision tree as the decision tree type for our Random Forest Model. By comparing the predicted data of ARIMA and Random Forest, we choose ARIMA Model to predict data. We also construct a VAR-Turtle Trading Strategy Model to construct our trading strategy. First, we use the turtle trading strategy to build our strategy pool. Then We use VaR values to replace the traditional ATR to measure the risk exposure of total assets, as VaR can have a more precise measurement of the risk of potential losses. Besides, we build Dynamic-Markowitz Model to make a comparison with VAR-Turtle Model. Moreover, we also construct a Greedy-Trading Model to make a comparison with VAR-Turtle Model. As a result, we find that VAR-Turtle Model aims to achieve the best asset arrangement based on the predicted price offered by ARIMA.