Stock Market Value Prediction Using Attention Based Long Short-Term Memory Model

Stock market value prediction is a trendy field these days with a lot of promise and challenges. An excess existing of methods for forecasting stock values have surfaced in recent times and the forecast accuracy of these systems is still far from optimal. This paper presents an Attention-Long Short- Term Memory model (Attention-LSTM) for stock market value prediction. The internet is the primary source of stock market data which offers historical data as well as technical indications such as stock posts and financial news. Next, combine historical and technical pointer's data to create a feature set for stock market value prediction, then utilize the Attention-LSTM network to make predictions which shows superior performance. The proposed model produces the better result of Mean Absolute Error (0.0479), Mean Squared Error (0.0050), Root Mean Squared Error (4.04), and Mean Absolute Percentage Error (1.0861) compared to previous models like Correlation Coefficient-Broad Learning System (PCC-BLS), Deep Learning-GuesSentiment Analysis (DL-GuesS), and LSTM.