Deep Learning Models for High Accuracy Stock Price Forecasting

Time-based sequence forecasting has been utilized in a wide variety of contexts to estimate future stock prices. The investigation and design stages of a financial model may significantly influence the decisions made by investors in order to ensure a successful trading process. Moreover, these types of forecasting methodologies are implemented to ascertain the selling prices of the stock. Nonetheless, as a consequence of the time series' non-linearity, there exists a possibility that the stock prices may be negatively impacted. In order to tackle these concerns, our intention is to integrate and prioritize the Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) variants of Recurrent Neural Networks. The model is undergoing training using Amazon stock price data spanning August months of 2013 to 2022, in addition to exploratory data analysis. Following an exhaustive performance analysis, it was determined that our intelligent time series-based GRU/LSTM forecasting system produced accurate results and was also robust with regard to nonlinear time series pertaining to the prediction of the closing prices of two stocks.