The dissemination of HIV/AIDS-related information on social media, particularly Twitter, plays a vital role in shaping public perception and increasing societal awareness. The vast and diverse nature of such data requires effective analytical techniques to interpret the sentiments expressed in tweets. This study proposes a deep learning-based sentiment classification approach using hybrid CNN-BiGRU and BiGRU- CNN models. These models combine the spatial feature extraction capabilities of Convolutional Neural Networks (CNN) with the contextual understanding of Bidirectional Gated Recurrent Units (BiGRU). A Genetic Algorithm (GA) is employed to optimize the model’s hyperparameters and enhance overall performance. Sentiments are categorized into positive, neutral, and negative classes. The model’s effectiveness is measured using the F1- score as the primary evaluation metric. Experimental results show that the GA-optimized BiGRU-CNN model achieves the best performance, reaching an F1-score of 85.70% in the final scenario. These findings confirm the model’s effectiveness in sentiment classification of HIV/AIDS-related Twitter data and highlight its potential to support public health communication strategies.
Keywords—Sentiment analysis, CNN-BiGRU, Twitter, HIV/AIDS, deep learning, text classification Introduction.