Accurately predicting stock market trends is essential for making informed financial decisions, helping investors minimise potential risks and refine their strategies. However, due to unforeseen market volatility and uncertainties caused by global events like the COVID-19 pandemic, this task has become increasingly challenging. This analysis focuses on two advanced models: Long Short-Term Memory (LSTM) and Spatio-Temporal Attention Neural Network (STANN), which are used to forecast stock trends. LSTM networks excel at capturing long-term patterns in time-series data, making them an effective tool for predicting future market behaviour. Additionally, the STANN model introduces spatio-temporal attention mechanisms, offering a novel approach to capturing both spatial and temporal dependencies in stock data. By analysing historical stock price data and their responses to market fluctuations, particularly those resulting from the economic impact of the COVID-19 crisis, our evaluation assesses the performance of these models. The results that we got were quite promising for both the LSTM and STANN models. The MAE calculated using the LSTM model for the Non-COVID (2010-2019) periods were 0.68, 0.52, and 1.2 for the stocks of Apple, Google, and Microsoft respectively. Additionally, the MAE for the COVID period (2010-2024) using the same model were 1.24, 1.43, and 2.76. On the other hand, the MAE values for the stocks using the STANN model were 3.846, 4.436, and 7.68 for the Non-COVID period and 8.077, 11.97, and 18.085 for the COVID period. The results underscore the importance of accurate stock market predictions, especially in times of economic uncertainty caused by the COVID-19 pandemic, and demonstrates the effectiveness of both the LSTM and STANN models in this domain.