In the present age, the e-commerce industry is experiencing rapid growth, with most people using e-commerce websites to buy products. User reviews are essential in this context, serving as a crucial element for sentiment analysis. This paper proposes a two-layered approach consisting of a base learner that incorporates a Deep Hybrid Parallel Fused CNN-LSTM (HPFCL) model, with CNN and LSTM models connected in parallel with a fused layer. The features obtained from the HPFCL model are combined through a fusion mechanism. The fused features are then inputted into meta-learner, which includes a machine learning classifier, to enable accurate classification. This model is unique in two main aspects: firstly, its application to analyzing Amazon reviews; secondly, its empirical evidence showcases superior performance compared to other models. This hybrid model is evaluated using three pre-trained word embedding techniques: GloVe, FastText, and Word2Vec. The study’s findings indicate that the HPFCL model outperformed other deep learning models across all three embeddings, highlighting its potential to enhance performance in sentiment analysis tasks.