Machine Learning-Based Election Results Prediction Using Twitter Activity

Elections are a vital component of democracy because they allow citizens to exercise their right to vote and choose representatives at all levels of government. The exponential growth of social media platforms has facilitated more accessible communication between political parties, politicians, and voters, disseminating distinctive political messages. Twitter, one of the most liked social media platforms, has become a valuable source for real-time public opinion. The proposed work uses sentiment analysis to predict electoral outcomes. Through analyzing the sentiment expressed in tweets regarding political candidates or parties, the machine learning algorithm, Long Short-Term Memory (LSTM), can be trained to identify patterns and correlations between sentiment and election results. The work examines the efficacy of the suggested approach by analyzing two large datasets of tweets during a specific election period. The proposed work presents a hybrid multimodal framework that incorporates emojis and multilingual text, including Hindi, English, and Hinglish (tweets containing both Hindi and English). The model is validated on datasets containing these multilingual elements and additionally tested on English-language tweets to evaluate its performance across different linguistic contexts. The results demonstrate the predictive capability of sentiment analysis on Twitter activity, shedding light on its potential as a supplementary tool for election forecasting. The outcome of the proposed method overviews Rahul Gandhi and Narendra Modi's polarity and subjectivity values. The accuracy of the proposed work is 87.65%, which outperforms the Twitter dataset used in state-of-art work.