Predicting Elections: Unveiling the Role of Social Media Data and Sentiment Analysis with ResBiLSTM Methodology

Abstract

Several volumetric, sentiment, and social network methodologies are presented and assessed in this study with the aim of predicting significant decisions from social media platforms. The opinions of the people are quite essential when it comes to determining certain major decisions. People from all walks of life have had a global platform for public expression on social media for almost twenty years. When trying to gauge public opinion, sentiment analysis also called opinion mining can be useful. Each step of the process from selecting a method to preprocessing, feature extraction, and training the model requires exact sequentially. There were two distinct phases to the data preparation process. Separate from cleaning the Twitter data that was retrieved is data wrangling, this comprises changing the format of the data. A feature extraction process makes use of ngrams, which are collections of n words taken from a source text. Extracting features is the first step in training a ResBiLSTM model. Both Resnet and BiLSTM, two state-of-the-art algorithms, are surpassed by this novel approach. The results show that the accuracy reached 95.45%, which is a huge improvement.