Adaptability of Financial Time Series Prediction Based on BiLSTM

Accurate prediction of financial market can promote the steady development of financial market, but the high frequency and high noise of financial time series make accurate prediction a challenging task. In this paper, bidirectional long short-term memory neural network (BiLSTM) in deep learning is applied in financial time series, and BiLSTM has one more layer of reverse structure to try to mine more effective information. The prediction performances of unidirectional long short-term memory neural network (LSTM), support vector regression (SVR) and differential autoregressive moving average model (ARIMA) are compared. The results show that BiLSTM model has the highest prediction accuracy, which can fully capture the past and future data information simultaneously, take the reverse relationship of data into account, and predict the long-term and short-term dynamic trends of financial time series effectively.