A Hybrid Deep Learning Model for Enhanced Customer Sentiment Analysis forE-commerce Platforms

Customer sentiment analysis has become more important for e-commerce systems if one wants to grasp consumer opinions and increase user happiness. This work proposes a new hybrid deep learning model combining CNN, BiLSTM, and Transformer layers in order to improve the bar for E-commerce review sentiment analysis. With its self-attention mechanism, the transformer layer provides context knowledge; the convolutional neural network (CNN) layer recognises textual local features; the long short-term memory (BLSTM) layer identifies sequential dependencies. Customer evaluations sometimes include sarcasm, uncertainty, and context changes; our hybrid architecture uses the strengths of each component to control these problems. This work uses the NFT dataset available on 9nftman.com. Test on a large dataset of online purchasing reviews, the model exceeded conventional models in terms of accuracy, F1-score, and recall rates. Crucially for a full knowledge of the customer, this improved sentiment prediction indicates that the computer can detect complex emotions including mixed or neutral input. Though the hybrid model raises processing complexity, its improved prediction accuracy and context sensitivity make it ideal for large-scale sentiment analysis in dynamic E-commerce contexts. Future study on model optimisation and interpretability will concentrate on finding a balance between computational efficiency and commercial usability. Accurate and context-aware sentiment analysis provides E-commerce platforms with strong new tools for learning about consumer experiences and making educated strategic decisions based on the proposed paradigm taken whole.