Predicting Stock Closing Price After COVID-19 Based on Sentiment Analysis and LSTM

Stock market prediction remains a challenging problem in the economic field because of its highly stochastic nature which is compounded by the uncertainty that the COVID-19 pandemic presents. To address the volatility and noise, we utilize a deep learning stock market prediction model based on investors’ sentiments, which allows us to better understand the direction the market will shift, improving the model prediction accuracy. We adopt an LSTM recurrent neural network because of its advantages in analyzing time-series data through its memory function and utilize an attention mechanism to focus the model on critical information. Experimental data demonstrate that utilizing the investors’ emotional tendencies and the attention mechanism can help LSTM to predict closing prices despite the uncertainty of the pandemic. Accurate predictions will allow shareholders and investors to understand market forces and emphasize sustainable investing and development.