The goal of investors investing in volatile assets is to maximize the growth of their asset value. Gold and bitcoin are both volatile assets. The best strategy to make profits from investing is to buy low and sell high. Therefore

We are committed to developing a traditional model.For part 1

First

the decision we make on a given day are based on what we think is going to happen that day. Prediction Model We use Arima sequence prediction model and SVM model to predict the trend of the data in the past and later days of the day

and use particle swarm optimization algorithm to optimize the most reasonable decision according to the trend. At the same time

we also use deep reinforcement learning

which is based on the data trend chart predicted by LSTM to make agents perform actions

keep learning

and finally achieve good results. At the same time

we modify the value of commission to analyze the sensitivity of the model.For part 2: we use the dynamic programming algorithm to prove that our model is the best. Comparing the ideal total return with the optimal total return of our model

it is found that the optimal total return of the model is close to more than two-thirds of the ideal total return. For part 3: we explore the impact of gold commission rate and bitcoin commission rate on the final total return. We found that the commission rate of bitcoin has a greater impact on the maximum return than that of gold

and the higher the commission rate

the smaller the return.