The stock market is an environment where prices fluctuate due to various factors such as economic conditions and global events, making trend prediction challenging. Current stock market bots, such as StockHero, rely on simple machine learning techniques like linear regression, which can capture short-term patterns but fail to model long-term dependencies and complex behaviors. Long Short-Term Memory (LSTM) networks, a type of deep learning model, perform well with sequential data and uncover underlying patterns, but they tend to overfit and struggle with short-term changes. The proposed Stock Prediction and Analysis Model (SPAM) integrates the strengths of LSTM networks and XGBoost, combining the former’s ability to model long-term dependencies with the latter’s effectiveness in capturing short-term patterns. SPAM was validated using datasets from Bank of China, Amazon, TCS, and Google, and experiments were conducted on a Lenovo Ideapad with an Intel Core i7 processor, 16GB RAM, and NVIDIA GTX 1650 GPU, as well as Google Colab with T4 GPUs. The results show that SPAM significantly outperforms existing models, achieving a reduction in mean squared error by 0.0004%, root mean squared error by 0.00191%, mean absolute error by 0.0004%, and an increase in R-squared value by 0.032%. This research offers valuable applications in financial decision-making, investment strategies, and automated trading systems, equipping investors and analysts with a robust tool to navigate volatile markets by providing both short-term insights and long-term trend predictions.