ABSTRACT
This nation-shaping election of 2020 plays a vital role in shaping the future of the U.S. and the entire world. With the growing importance of social media, the public uses them to express their thoughts and communicate with others. Social media have been used for political campaigns and election activities, especially Twitter. The researchers intend to predict presidential election results by analyzing the public stance toward the candidates using Twitter data. Previous researchers have not succeeded in finding a model that simulates well the U.S. presidential election system. This manuscript proposes an efficient model that predicts the 2020 U.S. presidential election from geo-located tweets by leveraging the sentiment analysis potential, multinomial naive Bayes classifier, and machine learning. An extensive study is performed for all 50 states to predict the 2020 U.S. presidential election results led by the state-based public stance for electoral votes. The general public stance is also predicted for popular votes. The true public stance is preserved by eliminating all outliers and removing suspicious tweets generated by bots and agents recruited for manipulating the election. The pre-election and post-election public stances are also studied with their time and space variations. The influencers' effect on the public stance was discussed. Network analysis and community detection techniques were performed to detect any hidden patterns. An algorithm-defined stance meter decision rule was introduced to predict Joe Biden as the President-elect. The model's effectiveness in predicting the election results for each state was validated by the comparison of the predicted results with the actual election results. With a percentage of 89.9%, the proposed model showed that Joe Biden dominated the electoral college and became the winner of the U.S. presidential election in 2020.
ABSTRACT
An efficient fake news detector becomes essential as the accessibility of social media platforms increases rapidly. Previous studies mainly focused on designing the models solely based on individual data sets and might suffer from degradable performance. Therefore, developing a robust model for a combined data set with diverse knowledge becomes crucial. However, designing the model with a combined data set requires extensive training time and sequential workload to obtain optimal performance without having some prior knowledge about the model's parameters. The presented study here will help solve these issues by introducing the unified training strategy to have a base structure for the classifier and all hyperparameters from individual models using a pretrained transformer model. The performance of the proposed model is noted using three publicly available data sets, namely ISOT and others from the Kaggle website. The results indicate that the proposed unified training strategy surpassed the existing models such as Random Forests, convolutional neural networks, and long short-term memory, with 97% accuracy and achieved the F1 score of 0.97. Furthermore, there was a significant reduction in training time by almost 1.5 to 1.8 × by removing words lower than three letters from the input samples. We also did extensive performance analysis by varying the number of encoder blocks to build compact models and trained on the combined data set. We justify that reducing encoder blocks resulted in lower performance from the obtained results.
