Integrating Swarm Intelligence with Deep Learning for Enhanced Social Media Sentiment Analysis
Parminder Singh, Saurabh Dhyani · International Journal of Electrical and Electronics Engineering · 2025
In our digital world, it is vital to know the opinion of the masses on social network sites. The study has presented a new framework to merge the advantages of deep learning and swarm intelligence, PSAM (Particle Swarm-Accelerated Model). It combines Long Short-Term Memory (LSTM) to perform a good sentiment classification and Particle Swarm Optimization (PSO) to perform good feature selection and hyperparameter optimization. The model has been used to label the YouTube reviews of the film Housefull 5 by classifying these sentiments as positive, negative or neutral with an impressive accuracy of 95.2 percent. The sentiment analysis pipeline starts with the extraction of comments through the YouTube API, and their pre-processing consists of removing punctuation, removing low-frequency words, normalising colloquial vocabulary, and emoji analysis. PSO has been essential in ensuring that the relevance of subsets has been determined and that feature reduction enhances the performance of the LSTM model tremendously. Comparing it with the conventional algorithms like Naïve Bayes, SVM, CNN and Random Forest, PSAM gives desirable results in all the major criteria of the build, like the accuracy and F1-score. The hybrid approach is very competent in sentiment analysis and has great potential to be extended to real-time systems and cross-platform social media mining.