Twitter Sentiment Analysis Using Clustering based Cuckoo Search Algorithm and Ensemble Classifier

In recent times, the sentiment analysis is an emerging research topic in data mining and Data Science. The traditional sentiment analysis based on machine learning methods focused only on text feature representations for constructing vector representations of the documents. The existing methods has rich semantic and grammatical data, but fails in extracting the sentiment data, which is essential for sentiment analysis. The metaheuristic-based clustering algorithms are effective compared to the traditional methods in sentiment analysis, because of the changing nature of the twitter datasets. In this manuscript, the Clustering based Cuckoo Search Algorithm (CCSA) is proposed along with ensemble classifier for sentiment analysis. After collecting the twitter data from twitter Sanders apple 2 dataset and twitter Sanders apple 3 dataset, the data pre-processing is performed for better data understandability. Then, five features such as positive expletive, negative expletive, profound words, absolute properties, and positive words, negative words and neutral words are practiced for mining the denoised data's feature vectors. The mined feature vectors are higher dimensions, so the CCSA is implemented for selecting the optimal feature vectors that reduces the system complexity, and enhances the computational time of the system. Finally, the contribution to the ensemble classifier for sentiment analysis is provided by the discriminative feature vectors. Simulation outcome exhibited that the ensemble-based CCSA model attained 93.45% and 93.85% of accuracy on the twitter-sanders apple 2 and twittersanders apple 3 datasets, where the attained experimental results are effective related to the existing models.