Sentiment analysis (SA) is a critical system that helps the understanding of the opinion of a large population of people, and it makes it easier to know the trend of the behavior on social sites. The proposed research work presents a hybrid intelligent method called PSAM (Particle Swarm-Accelerated Model), which combines Particle Swarm Optimization (PSO) to execute the best feature selection and Long Short-term Memory (LSTM) networks based on the deep learning approach to optimally identify the sentiment. With the integration of these two methods, the accuracy and the effectiveness of sentiment prediction in the presented model increase. Applying the model to real-world YouTube movie reviews, it would include considerable amount of preprocessing, including normalization of emojis, deletion of stopwords and removal of slang. PSO part brings more precision due to the isolation of the most beneficial parameters, and LSTM is a proficient examiner of the contextual flow in sequential data. In terms of accuracy PSAM has a remarkable performance of 96.2 while when compared to other traditional models such as SVM, CNN and Naive Bayes, the PSAM outperforms these traditional models. Such findings indicate the future potential of PSAM as an effective and robust method of sentiment analysis on any social media platform.