Robust Portfolio Optimization using LSTM-based Stock and Cryptocurrency Price Prediction: An Application of Algorithmic Trading Strategies

Algorithmic trading (AT) has become widely used recently because of its high speed and accuracy in implementing diverse and complex strategies. Using algorithms also allows traders to execute their trading strategies in a high volume and numerous transactions without involving human emotions. While AT has many advantages, it also carries some risks due to the uncertain stock market conditions and the impact of news and political, social, and other events. Therefore, forming a stock portfolio and stabilizing against uncertainties, in conjunction with accurate market predictions, can significantly reduce risk.in this paper, For the first time, we developed a robust portfolio optimization model based on LSTM prediction using the AT strategies based on short-term moving average techniques. First, we implement the strategies derived from the VLMA, FLMA, EMA, and SMA algorithms based on the LSTM's predicted price. Secondly, we develop a robust portfolio optimization model using the abovementioned algorithms. The results show that in both stock and crypto portfolios, moving average strategies will perform better than the benchmark strategy (Buy-and-hold). Also, when the model parameters are deterministic, the robust portfolio constructed stocks and crypto will perform better than Buy-and-hold for all algorithms. However, when the variance from certain models increases, VLMA and FLMA (۱۵-day holding) for stocks and FLMA (۳۰-day holding) for the crypto will not be a suitable investment option. Additionally, portfolios constructed using all AT strategies and all assets outperform the benchmark portfolio in certain and non-certain markets.

کلیدواژه ها:

Robust Portfolio Optimization ، Algorithmic trading ، price prediction ، LSTM