Abstract
Many modern computational social science academics are concerned about the recent trend of information sharing. Platforms for online social networks are being exploited to spread propaganda. These days, this is a lethal tool used to undermine democracies and other religious or political gatherings. Virtually every region of the planet was impacted by the political campaign. During the height of the 2024 Indian General Assembly election, A large volume of tweets has been shared that advocate various kinds of propaganda. The research proposes a cutting-edge strategy based on artificial neural networks that categorizes tweets into Propagandistic and Non-Propagandistic categories. Data is severed out with multiple ambiguous hashtags, and after the extraction, the approach is manually annotated into binary classes. Hybrid feature engineering has been used to merge the features of “Term Frequency (TF)/Inverse- Document Frequency (IDF), Tweet Length, and “Bag of Words,”. The planned approach is compared with logistic regression, SVMs, and multinomial naïve Bayes. The results showed that the Artificial-Neural-Network performed with a recall of 77% along with an accuracy of 77.15% and a precision of 79% in comparison with a number of other machine learning algorithms. A step forward could see deep learning techniques such as LSTM used for such classification assignments.
Keywords: Indian General Assembly Election, Machine Learning, Neural Networks, Online Social Network, Propaganda.