We refer to deep learning sentiment analysis of Amazon product reviews using with Graph Neural Networks (GNN) and Long Short-Term Memory (LSTM) units, as well as in this paper. In this research, we study the applicability of these models by dividing the reviews into three sentiment categories: neutral, negative and positive. It successfully captures both sequential dependencies and relational information by constructing a hybrid model that combines GNN and LSTM. The models are evaluated on a large dataset of Amazon reviews, and the hybrid model outperforms the individual GNN and LSTM models, achieving F1-score and the highest accuracy. The result shows that Hybrid (LSTM + GNN) is outer formed each other and achieved with the highest accuracy 87.3%. Highlight the importance of incorporating both sequential and relational data for accurate sentiment classification.