Exchange Rate Prediction Model with Bulls and Bears Game

In today's highly globalized economy, exchange rates play a crucial role. The fluctuation of price data makes exchange rate prediction a challenging problem. Existing research lacks the studies that integrate the continuous game between bulls and bears. We propose a prediction model for exchange rate based on Markov and Genetic LSTM. The game state sequence of bulls and bears is obtained from the fluctuation of price data, and Markov is used to extract the state transition features, which adds additional contextual information, allowing the model to better understand the evolutionary patterns between states. Considering the ability of LSTM to capture long-term dependencies of data, we further integrate state transition features into the training of LSTM. Next, LSTM is embedded into a genetic algorithm to obtain the optimal hyperparameters through iterative natural selection and evolution. Furthermore, using daily exchange rate data of the US dollar/Chinese yuan in the past year for verification, the experiment shows that the model has high accuracy and good generalization ability for medium-term exchange rate prediction.