Fake News Detection Using Machine Learning

Abstract

The domination of digital platforms for distribution of fake news traveling through social media outlets continues to be a pressing concern for sustained societal trust and the sanctity of information. This study focuses on evaluating the performances for different fake news detecting algorithms using machine learning techniques with the popular datasets “Fake.csv” and “True.csv”. In our study, we use of four algorithm models; LR, DTC, GBC and RFC. The data undergoes rigorous preprocessing stages, namely text cleansing, tokenization as well as TF-IDF vectorization. Furthermore, the data is then split into sets of training-test data, where model hyperparameter values are optimized utilizing the GridSearchCV framework. Accuracy, precision, recall and the F1-score calculated for these models prove the stated hypothesis where model Gradient Boosting Classifier outperforms the rest with the best metrics, while Logistic Regression showed good performance, which underlines the model’s usefulness in practice. The durability of DT and RF classifiers under rigorous tuning is noteworthy. The outcome of our study hold significant importance for the worlds of journalism, politics, and social media, stressing the significance of ML models in detecting fake news, thus enabling more public discourse and more informed debates. This study thus reiterates the need for ever- evolving techniques in machine learning to counter the issue of misinformation.