Comprehensive Analysis of Hybrid Machine Learning Models using Optimization Techniques for Improving Twitter Sentiment

Sentiment analysis (SA), an important part of natural language processing, is a key part of getting useful information from content generated by users, especially on social media sites like Twitter. This paper considers different techniques for analysing and improving sentiments, with a focus on the hybrid SVM_CNN model used with Twitter datasets. Support Vector Machines (SVMs) are used for classification, and Convolutional Neural Networks (CNNs) are used for feature extraction. The SVM_CNN model is proved to be an effective framework for emotion classification tasks. The study looks into different optimization methods, such as GridSearchCV, Genetic Algorithm, Particle Swarm Optimization (PSO), and Bayesian Optimization, to make the SVM_CNN model more accurate and reliable. The success of each optimization method is assessed by how successfully it determines the optimal hyper parameters that are specific to Twitter text data. The results show that Differential Evolution is the most accurate method, over other existing ones. This study helps to improve methods for sentiment analysis by showing how machine learning techniques and advanced optimization algorithms can work together to make them more useful.