This article mainly studies the application of fusion neural network model in the field of English sentiment analysis. The study first reviews the basic theory of sentiment analysis, the development of neural networks, and the concepts and principles of fusion models. Subsequently, this article deeply discusses text preprocessing technology, different sentiment analysis methods, and the application of convolutional neural network (CNN), recurrent neural network (RNN) and transformer network (Transformer) in sentiment analysis. In the experimental part, sentiment analysis solutions based on different models were designed. By comparing with the traditional single model, the advantages of the fusion model in terms of accuracy, stability and efficiency were demonstrated. The research results show that fusing different types of neural networks can effectively improve the accuracy of sentiment analysis, especially when dealing with complex and subtle emotional expressions. In addition, the paper also discusses the challenges of the model in terms of resource consumption and complexity, as well as possible future research directions, including cross-language sentiment analysis and computing resource optimization.