CNN-LSTM-Avg Fusion Network: A Sentiment Analysis Model Based on Multi-Feature Fusion

Sentiment analysis, as an important task in natural language processing, has been widely applied in multiple fields. This paper proposes a deep learning model for text sentiment analysis based on a three-way feature fusion of CNN, LSTM, and average pooling. The model achieves efficient identification of text sentiment by fusing the local semantic features captured by a Convolutional Neural Network (CNN), the sequence contextual information extracted by a Long Short-Term Memory network (LSTM), and the global semantic representation obtained through average pooling. Experiments were conducted on a public sentiment analysis dataset, and the results indicate that the proposed model achieved an accuracy of $88.20 \%$ on the sentiment classification task, outperforming various mainstream baseline models. Furthermore, ablation experiments validated the effectiveness of each component of the model, demonstrating that the fusion of the three types of features significantly enhances the model performance. This study provides a new multi-feature fusion framework for sentiment analysis, which holds certain reference value for research in related fields.