A Time Weighted Based Deep Learning Model for Stock Market Price Prediction

Stock price prediction has recently become a popular topic of research, with typical forecasting techniques based on statistical and econometric models. These models meet complications when dealing with nonstationary time series data. With the rapid expansion of the internet and the increasing popularity of social media, online news and comments have become useful sources of information into investor feelings and attitudes towards equities. These sources give crucial information for stock price prediction. In this research, the purpose is to construct an innovative stock price estimation algorithm by using social media data. To balance the dimensions between text feature variables and stock financial index variables, the study decreases the dimensionality of the text feature vectors using a Stacked Auto-Encoder (SAE). Ultimately, the research proposes a Time-Weighted Long Short-Term Memory (TW-LSTM) model to estimate stock values accurately. The experimental results reveal that this strategy exceeds all three benchmark models in various assessment measures and works well in predicting stock prices. This research offers a novel stock price prediction model that mixes conventional financial data with social media text features generated through deep learning technology.