Algorithm Fusion and Multi-Model Integration Method in Media Communication Effect Evaluation

The media communication effect refers to the influence that various forms of media, such as television, social media, radio, and digital platforms, have on individuals’ thoughts, emotions, behaviors, and attitudes. Understanding how the media affects public opinion, information dissemination, and social norms is essential for those studying communication studies, marketing, politics, and psychology, among other disciplines. This study uses an innovative approach that combines algorithm fusion and multi-model integration to enhance the evaluation process of media communication effects. However, constructing an opinion model effectively is severely restricted by the lack of labeled data. To address this issue, this study proposed a novel approach for public opinion sentiment analysis using multi-model fusion with the moth flame optimizer with refined long short-term memory (MFO-RLSTM) model. To improve the accuracy of sentiment identification, the attention mechanism is also included to concentrate on important emotional features. Sentiment analysis of public opinion is conducted using data from social media platforms like Twitter, Facebook, and Instagram, which collect user-generated content through images, text, and audio. This method improves sentiment feature learning efficiency by utilizing the benefits of multi-models, even while labeled data are scarce. The data were preprocessed using Noise removal and tokenization for obtained data. A convolutional neural network (CNN) is used to extract both local and global features. This fusion of local and global characteristics offers a thorough comprehension of public sentiment. Performance is assessed using F1-score, recall, accuracy, and precision. The findings show that the suggested approach significantly enhances sentiment analysis when applied to media communication.