A Comprehensive Analysis of Predicting Price Movement in Daily Stock Data

Abstract

Accurate prediction of stock price movements is a quite challenging, especially when it is about the financial sectors. To be very specific, the stock market, is volatile in most of the times. This research focuses on forecasting stock price movements for Square Pharmaceuticals PLC, a leading listed issuer company. It's listed in the Dhaka Stock Exchange PLC (DSE). The study emphasizes the use of daily stock data from the year 2017 to 2023. This study examines financial indicators like opening price, closing price, high, low, adjusted close, and trading volume as the key. The research employs advanced deep learning algorithms, specifically Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks, to capture temporal dependencies and non-linear patterns present in the data. The LSTM model is found to be more precise, producing lesser errors and higher R² statistics for the training set, while GRU converges at a higher speed and effectively captures short-term dependencies. However, both encountered challenges in extrapolating their predictions onto the test set: stock price forecasting presents inherent difficulties, especially in upcoming developing markets like Bangladesh. A coherent literature review on the recent advancements of stock market prediction inspired the study, which delves into the investor sentiment analysis, hybrid machine learning approach, and reinforcement learning. The integration of this information will contribute toward filling the existing knowledge cracks and provide practical recommendations to investors, analysts, and researchers who are looking for data-driven strategies for market analysis.