Design Stock Market Trading Strategy with Deep Learning: A Bi-LSTM Based Approach

For this paper, we utilize the famous LSTM model and modify it to a model that consists of two layers of Bi-LSTM. Our trading strategy with the model is to trade the stocks with the highest growth rates predicted by the model and the strategy repeat once a day. To test the efficiency of our model, we change parameters of the experiment: time-interval and both number and frequency of buying stocks. In this way, we can conclude that the model is stable. Whether reducing the frequency of buying increases profit depending on the degree of this parameter. And the number of buying stocks sometimes prevent greater loss during the strategy.