Innovative artificial intelligence approach to analyze political opinions from social media data
Year: 2025
Type: article
Abstract: The growing influence of social media platforms on political discourse has been reshaping the formation and expression of public opinion, particularly in relation to voting behavior. Social media has been recognized as a rich, real-time source of data, offering valuable insights into political sentiments, voting patterns, and electoral dynamics. Conventional techniques for political assessment analysis have had difficulty capturing the dynamic, varied, and quickly changing character of voter mood in online spaces. In this study, an innovative artificial intelligence (AI) approach was developed to analyze political opinions extracted from social media data, with a specific focus on predicting voting behavior. Data were gathered from social media sites including Facebook, Reddit, and Twitter, focusing on popular opinion and political conversations throughout significant events and election seasons. The data were pre-processed using tokenization techniques to improve model accuracy, and features were extracted through Term Frequency–Inverse Document Frequency (TF-IDF). A brand-new method called Intelligent Coot Bird Optimized Categorical Boosting (ICBO-CatBoost) was put forward to analyze political viewpoints in social media data. This method integrates the capabilities of the CatBoost algorithm—known for its handling of categorical variables—with the Intelligent Coot Bird Optimization algorithm, which was employed to fine-tune model hyperparameters and enhance performance in modeling complex, non-linear relationships within political discourse and voting trends. The ICBO-CatBoost model was shown to outperform traditional models, achieving an F1-score of 89%, accuracy of 92%, precision of 90%, and recall of 85%, thereby providing more reliable predictions of voting intentions. Furthermore, the model was able to capture subtle shifts in political sentiment over time, particularly surrounding key electoral events. These findings underscore the significance of leveraging social media data to predict voting trends and support political analysts, campaign strategists, and policymakers in better understanding electoral behavior.
Authors: Sumitra Menaria, Vivek Saraswat, Sidhant Das, Hemanth Kumar, Pooja Goel, Murari Devakannan Kamalesh
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4414814767