Distinguishing Truth from Deception: A Machine Learning Approach to Fake News Dectection

Abstract

The fast-paced dissemination of false information on social media is dangerous not only for public trust but also for political stability and societal cohesion. Tackling this issue, the current article is creating a machine learning framework for fake news identification, which is based on the Input-Process-Output (IPO) model to do the research work systematically. Together with the use of Natural Language Processing (NLP) tools, a couple of statistical feature validation, and supervised learning models, specifically, Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) networks, this research attempts to create stable interpretable, and reliable classification system. Trained and untrained news sources' textual data were gathered from which sentiment analysis, TF-IDF vectorization, and syntactic feature extraction were conducted as initial processing tasks. Statistical techniques such as Chi-square tests, T-tests, Pearson correlation coefficients, etc., were applied to pinpoint Feature 100 as the key attribute among the lot. The findings of the study reveal that the LSTM model significantly beat SVM in the case of class accuracy with a high precision and recall rate, which finally led to 94% of the students mastering the tests and obviously having broad skills. The research's main point is the fact that the union of statistical methods and deep learning models is necessary to make fake news detection much more effective. This study adds new knowledge about the making of a dependable automatic misinformation filtering system and further updates a safer digital information environment.