Forecasting stock prices with long-short term memory neural network based on attention mechanism

• DOI: 10.1371/journal.pone.0227222
• Publication Date: 2020-01-03T16:07:11Z
• Authors: Jiayu Qiu; Bin Wang; Changjun Zhou
• Keywords: Stock (firearms)

The stock market is known for its extreme complexity and volatility, people are always looking an accurate effective way to guide trading. Long short-term memory (LSTM) neural networks developed by recurrent (RNN) have significant application value in many fields. In addition, LSTM avoids long-term dependence issues due unique storage unit structure, it helps predict financial time series. Based on attention mechanism, a wavelet transform used denoise historical data, extract train features, establish the prediction model of price. We compared results with other three models, including model, denoising gated unit(GRU) network S&P 500, DJIA, HSI datasets. Results from experiments 500 DJIA datasets show that coefficient determination attention-based both higher than 0.94, mean square error our lower 0.05.