Covid-19 anti-vaccination detection from text data using feature selection approaches in deep learning

The COVID-19 pandemic has turned into a crisis that deeply affects health, economy and social life around the world. During this crisis, anti-vaccination poses a significant obstacle to the control of the epidemic and the effectiveness of vaccination campaigns. In this study, it was aimed to detect COVID-19 vaccine opposition from text data by using a combination of deep learning and feature selection approaches. The proposed method involves the integration of deep learning model and feature selection techniques and identifies anti-vaccine sentiment by identifying important features in text data. This study also focuses on evaluating the effectiveness of deep learning models. Among the hypotheses of the thesis, it is assumed that deep learning models can accurately classify and identify COVID-19 anti-vaccination by using them in an integrated manner with feature selection approaches. The sample used consists of English tweets about COVID-19, which are publicly available on GitHub and created using the Twitter API from social media platforms. The sample size included more than 10,000 text items to represent a wide range of data. As a method, different methods, deep learning model and optimization algorithms were used for feature selection. The proposed method involves the integration of deep learning model and feature selection techniques and identifies anti-vaccine sentiment by identifying important features in text data. TF-IDF and N-gram methods were integrated for feature extraction. The dataset consists of two labels. Since the labels are not balanced, the SMOTE method was applied to balance the dataset. Then, feature selection was performed using Chi-square. LSTM (Long Short Term Memory), one of the Deep Learning architectures, was used for the classification process. In classification analyses, the dataset is divided into 80% training and 20% test data. The results obtained with the proposed model were 99.23% accuracy value and 99.21% F1-score value. These results show that the proposed method can be successfully used to effectively detect COVID-19 vaccine opposition on text data. The results of the study can provide valuable information for the development of health policies and public information strategies. The findings of the study show that deep learning and feature selection methods can be used effectively to detect COVID-19 vaccine opposition. This research offers health authorities a powerful new tool to identify anti-vaccine sentiment when planning vaccination campaigns and designing public health interventions.