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

Abstract: Distribution of fake news on social media is so fast that it endangers public trust and confidence in the political system and society. This paper introduces a powerful machine learning system for the fake news detection that applies the Input-Process-Output (IPO) model for the systematic research procedure. Thus, based on Natural Language Processing (NLP), statistical feature validation, and supervised learning models—Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) networks—we have successfully built a classification system that is not only accurate but also interpretable and consistent. Words from both fake and verified news sources were processed using sentiment analysis, TF-IDF vectorization, and syntactic feature extraction. Such statistical features as Chi-square tests, T-tests, and Pearson correlation coefficients singled out "Sentiment Polarity Variance" (Feature 100) as the most important one among a number of the features. The SVM model was found to be less effective than the LSTM model as the latter reached 94% for the overall accuracy along with precision (0.93), recall (0.94), and F1-score (0.93). This study strengthens the argument that there is a possibility of a combination of statistical and deep learning being used for the purpose of detection and reducing misinformation and confirming the appearance of safer digital information ecosystems.