A Two-stage Model: Investment Market Trading Model Based on Prediction and Adaptive Strategies

Market traders buy and sell volatile assets frequently, with a goal to maximize their total return. Two such are gold bitcoin. This paper constructs two-stage model for price prediction trading strategy formulation. Firstly, we build dynamic ARIMA-LSTM hybrid model. And before applying it, use CEEMDAN method decompose the non-stationary time series first then reconstruct final result by predicting each IMF summing up weighted. can update training set dynamically when new data is released. After getting prices predicted, calculated several quantitative indicators so that make decisions more comprehensively instead of only focusing on predicted price. semi-supervised SVM develop an adaptive Finally, demonstrate superiority our from two perspectives. In actual investment transactions, be used as guide formulation strategies, thereby avoiding risks increasing returns.