Levy distribution-based Dung Beetle Optimization with Support Vector Machine for Sentiment Analysis of Social Media

The social media platforms such as Facebook and Twitter have increasingly popular to express their opinions on various topics. These platforms have emerged as efficient and largest source for collecting public opinion. Therefore, Levy distribution-based Dung Beetle Optimization (LDBO) with Support Vector Machine (SVM) is proposed for sentiment analysis. The LDBO enhances the exploration abilities which enables to search solution space effectively and avoids local optima. The SVM is efficient to find optimal hyperplane which reduces the margin among classes thereby leads classification performance. Moreover, SVM is useful with various kernel function which enables to capture non-linear relationship in data thereby enabling better generalization in unseen data. The Term Frequency-Inverse Component Frequency (TF-ICF) is helpful for differentiating among positive and negative sentiments. Moreover, it reduces the commonly occurred words and concentrates on informative terms. The preprocessing includes filtering, normalization, stemming, stop word removal and tokenization which enhances the quality and model performance. The SVM achieves better in terms of 98.65% accuracy, 98.12% precision, 97.36% recall and 97.73% f1-score compared to Gated Attention Recurrent Network (GARN).